{
  "version": "bekko.system_one.v1.renew.20260928",
  "s1mb_release_version": "s1mb.open_jev.curated.v2.20260928",
  "datasets": {
    "aegis2": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Exclude REDACTED placeholders; distinguish human and llm_jury labels; do not treat absent response labels as safe.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/nvidia/Aegis-AI-Content-Safety-Dataset-2.0/blob/d86bb8bedff51d25ac834ab7838f1cc61acb7a2c/README.md",
          "what_verified": "Upstream dataset/card identifier nvidia/Aegis-AI-Content-Safety-Dataset-2.0 at pinned revision d86bb8bedff51d25ac834ab7838f1cc61acb7a2c; license metadata is ['cc-by-4.0']. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/nvidia/Aegis-AI-Content-Safety-Dataset-2.0/raw/d86bb8bedff51d25ac834ab7838f1cc61acb7a2c/README.md",
          "what_verified": "Primary source card was fetched and read at d86bb8bedff51d25ac834ab7838f1cc61acb7a2c; card excerpt: # 🛡️ Nemotron Content Safety Dataset V2 <!-- Provide a quick summary of the dataset. --> The **Nemotron Content Safety Dataset V2**, formerly known as Aegis AI Content Safety Dataset 2.0, is comprised of `33,416` annotated interactions between humans and LLMs, split into `30,007` training samples, `1,445` validation samples,  and `1,964` test samples. This release is an extension of the previously published [Nemotron Content Safety Dataset V1](https://huggingface.co/datasets/nvidia/Aegis-AI-Cont"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/aegis2/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 3 source artifacts; 3 referenced local cache paths exist and integrity digests match. Plan source revision d86bb8bedff51d25ac834ab7838f1cc61acb7a2c."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/aegis2/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/aegis2",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 30331 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 9 actual rows from 3 pinned raw source files for source key aegis2; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "false": "The prompt does not violate the unsafe-content categories defined in the instruction. Content that only needs caution is not unsafe.",
            "true": "The prompt violates at least one unsafe-content category defined in the instruction."
          },
          "id": "aegis2__noul__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Source-grounded false/true meanings; see source evidence.",
          "selector": {
            "decision_id": "prompt_unsafe",
            "kind": "judgment",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The response does not violate the unsafe-content categories defined in the instruction. Content that only needs caution is not unsafe.",
            "true": "The response violates at least one unsafe-content category defined in the instruction."
          },
          "id": "aegis2__noul__001",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Source-grounded false/true meanings; see source evidence.",
          "selector": {
            "decision_id": "response_unsafe",
            "kind": "judgment",
            "type": "noul"
          }
        }
      ],
      "source_profile": "aegis2"
    },
    "laya__ag_news": {
      "disposition": "include",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching.",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "eb185aade064a813bc0b7f42de02595523103ca4",
          "sha256": "0ae9549054c5eca6a2284b1a6e8505b1ec92a39613aca402336429762b544a62",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/fancyzhx/ag_news/eb185aade064a813bc0b7f42de02595523103ca4/README.md?%2Fdatasets%2Ffancyzhx%2Fag_news%2Fresolve%2Feb185aade064a813bc0b7f42de02595523103ca4%2FREADME.md=&etag=%22fc01bffaf6b7ff9e9df5818a6fa2bd5cb4a27438%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "ag_news:default:test:0",
          "local_path": "output/typed_decisions/laya-v1/cases/cache/9c39e72ade56d84718626766e0b249d1edf6556888d66ed5dc56f085bf7e7bc2",
          "original_row_match_status": "matched",
          "sha256": "71de87ec66bc5737752a2502204dfa6d7fe9856ade3ea444dc6317789a4f13fb",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://huggingface.co/datasets/fancyzhx/ag_news/resolve/eb185aade064a813bc0b7f42de02595523103ca4/data/test-00000-of-00001.parquet"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/laya_adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "business": "Business and economic news.",
            "sci_tech": "Science and technology news.",
            "sports": "Sports news.",
            "world": "World news and international politics."
          },
          "id": "topic",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "Descriptions are copied from the pinned source conversion’s case criteria. Keep IDs aligned; only expand them from source definitions, never from the observed target.",
          "option_semantics": "fixed_source_taxonomy",
          "selector": {
            "decision_id": "topic",
            "type": "choice"
          }
        }
      ],
      "source_profile": "laya__ag_news"
    },
    "aqua_rat": {
      "disposition": "include",
      "disposition_reason": "Reviewed English typed judgment conversion; preserve selected source splits and label semantics.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/deepmind/aqua_rat/blob/33301c6a050c96af81f63cad5562cb5363e88971/README.md",
          "what_verified": "Upstream dataset/card identifier deepmind/aqua_rat at pinned revision 33301c6a050c96af81f63cad5562cb5363e88971; license metadata is ['apache-2.0']. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/deepmind/aqua_rat/raw/33301c6a050c96af81f63cad5562cb5363e88971/README.md",
          "what_verified": "Primary source card was fetched and read at 33301c6a050c96af81f63cad5562cb5363e88971; card excerpt: # Dataset Card for AQUA-RAT ## Table of Contents - [Dataset Description](#dataset-description) - [Dataset Summary](#dataset-summary)"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/aqua_rat/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 3 source artifacts; 3 referenced local cache paths exist and integrity digests match. Plan source revision 33301c6a050c96af81f63cad5562cb5363e88971."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/aqua_rat/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/aqua_rat",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 97675 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 9 actual rows from 3 pinned raw source files for source key aqua_rat; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "id": "aqua_rat__choice__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Preserve source Choice option IDs/descriptions and order; sampled option lists can be instance-specific.",
          "selector": {
            "decision_id": "answer",
            "kind": "judgment",
            "type": "choice"
          }
        }
      ],
      "source_profile": "aqua_rat"
    },
    "arc": {
      "disposition": "include_with_caveat",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching. Caveats: Existing source spot-check note: \"科学QAの選択肢とanswerKeyの対応を確認。Easy/Challengeの元splitを保持。\"",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "210d026faf9955653af8916fad021475a3f00453",
          "sha256": "4a293e352253d771604725c978adcea71ee6504e89ed40e1cb0024ecafa9b37e",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/allenai/ai2_arc/210d026faf9955653af8916fad021475a3f00453/README.md?%2Fdatasets%2Fallenai%2Fai2_arc%2Fresolve%2F210d026faf9955653af8916fad021475a3f00453%2FREADME.md=&etag=%22d50d3b885a6d04fe8cb16cc9947f64dff0ba941a%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "arc:ARC-Challenge:test:0",
          "local_path": "output/typed_decisions/v1/cache/616b0fd53ffc9ac51ef409f17c9aa3357c98b37d5283ab6eabcf2d0d37db01e6",
          "original_row_match_status": "matched",
          "sha256": "62f03257e737aed263f55c6abf87c7bb0028a44a6bdd2a26eb1279eb42c1d1e9",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://huggingface.co/datasets/allenai/ai2_arc/resolve/5a61ed188b8810abf32686fa83eb93787cbc5f25/ARC-Challenge/test/0000.parquet"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {},
          "id": "answer",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "Options and their descriptions vary by case; use the row’s own option-ID/description pairs. Do not assign a fixed meaning to option IDs across rows.",
          "option_semantics": "per_case_candidates",
          "preserve_case_options": true,
          "selector": {
            "decision_id": "answer",
            "type": "choice"
          }
        }
      ],
      "source_profile": "arc"
    },
    "arct": {
      "disposition": "include",
      "disposition_reason": "See permissive dataset build note; short task-specific instructions, source-specific label checks and grouped splits. Author license overrides missing mirror metadata where documented.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/tasksource/arct/blob/c13f83ea610d0f04c8b6ea50a59339b8204dcd44/README.md",
          "what_verified": "Upstream dataset/card identifier tasksource/arct at pinned revision c13f83ea610d0f04c8b6ea50a59339b8204dcd44; license metadata is ['apache-2.0']. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/tasksource/arct/raw/c13f83ea610d0f04c8b6ea50a59339b8204dcd44/README.md",
          "what_verified": "Primary source card was fetched and read at c13f83ea610d0f04c8b6ea50a59339b8204dcd44; card excerpt: # The Argument Reasoning Comprehension Task: Identification and Reconstruction of Implicit Warrants https://github.com/UKPLab/argument-reasoning-comprehension-task ```bib @InProceedings{Habernal.et.al.2018.NAACL.ARCT, title     = {The Argument Reasoning Comprehension Task: Identification"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/arct/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 3 source artifacts; 3 referenced local cache paths exist and integrity digests match. Plan source revision c13f83ea610d0f04c8b6ea50a59339b8204dcd44."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/arct/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/arct",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 1970 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 9 actual rows from 3 pinned raw source files for source key arct; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "id": "arct__choice__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Preserve source Choice option IDs/descriptions and order; sampled option lists can be instance-specific.",
          "selector": {
            "decision_id": "warrant",
            "kind": "judgment",
            "type": "choice"
          }
        }
      ],
      "source_profile": "arct"
    },
    "argument_quality": {
      "disposition": "include_with_caveat",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching. Caveats: Existing source spot-check note: \"WAの重み付き推奨確率とstanceを別判断として確認。文法の悪い低品質argumentでも支持/反対が別に存在する。WAは単純票割合と偽らずweighted_annotation_probability、MACE-Pも保持。\"",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "590726b3765b1b90c5e53a17e3b1f77d92d3aa8a",
          "sha256": "fc84df32b27dd27d62c1daa2fd8716ccb66b8ed7e13130a77a28e2a52b5eb700",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/ibm-research/argument_quality_ranking_30k/590726b3765b1b90c5e53a17e3b1f77d92d3aa8a/README.md?%2Fdatasets%2Fibm-research%2Fargument_quality_ranking_30k%2Fresolve%2F590726b3765b1b90c5e53a17e3b1f77d92d3aa8a%2FREADME.md=&etag=%22b8caf713190be46277c3726054293833d6fd6710%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "argument_quality:argument_quality_ranking:test:0",
          "local_path": "output/typed_decisions/v1/cache/80d73af733668383863280a534acc80957708021017d05bed5d97492db0cb022",
          "original_row_match_status": "matched",
          "sha256": "16c1c7d5309c663d26455726be03d5d704844aba87eaeabad446b02bd1367c00",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://huggingface.co/datasets/ibm-research/argument_quality_ranking_30k/resolve/977b0042dc91e5400d1ebdd71a27d84bcc37c3a8/argument_quality_ranking/test/0000.parquet"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/extra_adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "against": "The argument is against the topic.",
            "for": "The argument supports the topic.",
            "neutral": "The argument takes no position on the topic."
          },
          "id": "stance",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "Descriptions are copied from the pinned source conversion’s case criteria. Keep IDs aligned; only expand them from source definitions, never from the observed target.",
          "option_semantics": "fixed_source_taxonomy",
          "selector": {
            "decision_id": "stance",
            "type": "choice"
          }
        },
        {
          "criteria_by_id": {
            "false": "The argument as written is not suitable for a friend to use as is in that speech.",
            "true": "The argument as written is suitable for a friend to use as is in a speech either supporting or contesting the stated topic, regardless of the listener’s personal opinion."
          },
          "id": "recommend",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "",
          "selector": {
            "decision_id": "recommend",
            "type": "noul"
          }
        }
      ],
      "source_profile": "argument_quality"
    },
    "babi_nli": {
      "disposition": "include",
      "disposition_reason": "Reviewed English typed judgment conversion; preserve selected source splits and label semantics.; For entailment-style Noul rows, false means the queried support/entailment relation is not established; do not relabel all false examples as contradictions.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/tasksource/babi_nli/blob/80b689dc668b50f3640fc5368fd1893d385cb4e1/README.md",
          "what_verified": "Upstream dataset/card identifier tasksource/babi_nli at pinned revision 80b689dc668b50f3640fc5368fd1893d385cb4e1; license metadata is ['bsd']. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/tasksource/babi_nli/raw/80b689dc668b50f3640fc5368fd1893d385cb4e1/README.md",
          "what_verified": "Primary source card was fetched and read at 80b689dc668b50f3640fc5368fd1893d385cb4e1; card excerpt: # bAbi_nli bAbI tasks recasted as natural language inference. https://github.com/facebookarchive/bAbI-tasks"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/babi_nli/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 60 source artifacts; 60 referenced local cache paths exist and integrity digests match. Plan source revision 80b689dc668b50f3640fc5368fd1893d385cb4e1."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/babi_nli/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/babi_nli",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 34994 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 180 actual rows from 60 pinned raw source files for source key babi_nli; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "false": "The story does not support the hypothesis under the reasoning convention stated in the instruction.",
            "true": "The story supports the hypothesis under the reasoning convention stated in the instruction."
          },
          "id": "babi_nli__noul__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Source-grounded false/true meanings; see source evidence.",
          "selector": {
            "decision_id": "entailed",
            "kind": "judgment",
            "type": "noul"
          }
        }
      ],
      "source_profile": "babi_nli"
    },
    "banking77": {
      "disposition": "include_with_caveat",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching. Caveats: Existing source spot-check note: \"77意図の対応は良好。ただし「アプリを再インストールすれば解決するか」のように本文だけでは意図を断定しにくい元例がある。元ラベルは変更しない。\"",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "90d4e2ee5521c04fc1488f065b8b083658768c57",
          "sha256": "b494ae99274ad98adfedfefcb504469197a8732f9726f678320f4aaafc5502bf",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/PolyAI/banking77/90d4e2ee5521c04fc1488f065b8b083658768c57/README.md?%2Fdatasets%2FPolyAI%2Fbanking77%2Fresolve%2F90d4e2ee5521c04fc1488f065b8b083658768c57%2FREADME.md=&etag=%224596e6a6805b60ced7b681c6174c04dc9c90bec5%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "banking77:default:test:0",
          "local_path": "output/typed_decisions/v1/cache/6b8be95ee381517e269218a9b809a9c79da9d6796dd4d580b2f81db3efd48d9f",
          "original_row_match_status": "matched",
          "sha256": "d12d6e3bc4c3103966ae786dc435913c0c563dfa328f5a3646d0e62cfeeb474d",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://raw.githubusercontent.com/PolyAI-LDN/task-specific-datasets/57ec275d8078af65b7731c2a98be812d844a6d6b/banking_data/test.csv"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "Refund_not_showing_up": "Refund not showing up",
            "activate_my_card": "activate my card",
            "age_limit": "age limit",
            "apple_pay_or_google_pay": "apple pay or google pay",
            "atm_support": "atm support",
            "automatic_top_up": "automatic top up",
            "balance_not_updated_after_bank_transfer": "balance not updated after bank transfer",
            "balance_not_updated_after_cheque_or_cash_deposit": "balance not updated after cheque or cash deposit",
            "beneficiary_not_allowed": "beneficiary not allowed",
            "cancel_transfer": "cancel transfer",
            "card_about_to_expire": "card about to expire",
            "card_acceptance": "card acceptance",
            "card_arrival": "card arrival",
            "card_delivery_estimate": "card delivery estimate",
            "card_linking": "card linking",
            "card_not_working": "card not working",
            "card_payment_fee_charged": "card payment fee charged",
            "card_payment_not_recognised": "card payment not recognised",
            "card_payment_wrong_exchange_rate": "card payment wrong exchange rate",
            "card_swallowed": "card swallowed",
            "cash_withdrawal_charge": "cash withdrawal charge",
            "cash_withdrawal_not_recognised": "cash withdrawal not recognised",
            "change_pin": "change pin",
            "compromised_card": "compromised card",
            "contactless_not_working": "contactless not working",
            "country_support": "country support",
            "declined_card_payment": "declined card payment",
            "declined_cash_withdrawal": "declined cash withdrawal",
            "declined_transfer": "declined transfer",
            "direct_debit_payment_not_recognised": "direct debit payment not recognised",
            "disposable_card_limits": "disposable card limits",
            "edit_personal_details": "edit personal details",
            "exchange_charge": "exchange charge",
            "exchange_rate": "exchange rate",
            "exchange_via_app": "exchange via app",
            "extra_charge_on_statement": "extra charge on statement",
            "failed_transfer": "failed transfer",
            "fiat_currency_support": "fiat currency support",
            "get_disposable_virtual_card": "get disposable virtual card",
            "get_physical_card": "get physical card",
            "getting_spare_card": "getting spare card",
            "getting_virtual_card": "getting virtual card",
            "lost_or_stolen_card": "lost or stolen card",
            "lost_or_stolen_phone": "lost or stolen phone",
            "order_physical_card": "order physical card",
            "passcode_forgotten": "passcode forgotten",
            "pending_card_payment": "pending card payment",
            "pending_cash_withdrawal": "pending cash withdrawal",
            "pending_top_up": "pending top up",
            "pending_transfer": "pending transfer",
            "pin_blocked": "pin blocked",
            "receiving_money": "receiving money",
            "request_refund": "request refund",
            "reverted_card_payment?": "reverted card payment?",
            "supported_cards_and_currencies": "supported cards and currencies",
            "terminate_account": "terminate account",
            "top_up_by_bank_transfer_charge": "top up by bank transfer charge",
            "top_up_by_card_charge": "top up by card charge",
            "top_up_by_cash_or_cheque": "top up by cash or cheque",
            "top_up_failed": "top up failed",
            "top_up_limits": "top up limits",
            "top_up_reverted": "top up reverted",
            "topping_up_by_card": "topping up by card",
            "transaction_charged_twice": "transaction charged twice",
            "transfer_fee_charged": "transfer fee charged",
            "transfer_into_account": "transfer into account",
            "transfer_not_received_by_recipient": "transfer not received by recipient",
            "transfer_timing": "transfer timing",
            "unable_to_verify_identity": "unable to verify identity",
            "verify_my_identity": "verify my identity",
            "verify_source_of_funds": "verify source of funds",
            "verify_top_up": "verify top up",
            "virtual_card_not_working": "virtual card not working",
            "visa_or_mastercard": "visa or mastercard",
            "why_verify_identity": "why verify identity",
            "wrong_amount_of_cash_received": "wrong amount of cash received",
            "wrong_exchange_rate_for_cash_withdrawal": "wrong exchange rate for cash withdrawal"
          },
          "id": "intent",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "Descriptions are copied from the pinned source conversion’s case criteria. Keep IDs aligned; only expand them from source definitions, never from the observed target.",
          "option_semantics": "fixed_source_taxonomy",
          "selector": {
            "decision_id": "intent",
            "type": "choice"
          }
        }
      ],
      "source_profile": "banking77"
    },
    "bbq": {
      "disposition": "include_with_caveat",
      "disposition_reason": "See permissive dataset build note; short task-specific instructions, source-specific label checks and grouped splits. Author license overrides missing mirror metadata where documented.; No supervised training split exists in this release subset; treat available rows as evaluation-only/source-limited.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/lighteval/bbq_helm/blob/10937569dd76b6f33b2ff3572382ed87458e7856/README.md",
          "what_verified": "Upstream dataset/card identifier lighteval/bbq_helm at pinned revision 10937569dd76b6f33b2ff3572382ed87458e7856; license metadata is ['cc-by-4.0']. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/lighteval/bbq_helm/raw/10937569dd76b6f33b2ff3572382ed87458e7856/README.md",
          "what_verified": "Primary source card was fetched and read at 10937569dd76b6f33b2ff3572382ed87458e7856; card excerpt: --- dataset_info: - config_name: Age features: - name: context dtype: string - name: question dtype: string"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/bbq/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 1 source artifacts; 1 referenced local cache paths exist and integrity digests match. Plan source revision 10937569dd76b6f33b2ff3572382ed87458e7856."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/bbq/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/bbq",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 1000 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 3 actual rows from 1 pinned raw source files for source key bbq; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "id": "bbq__choice__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Preserve source Choice option IDs/descriptions and order; sampled option lists can be instance-specific.",
          "selector": {
            "decision_id": "answer",
            "kind": "judgment",
            "type": "choice"
          }
        }
      ],
      "source_profile": "bbq"
    },
    "boardgameqa": {
      "disposition": "include",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching.",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "78e38c3c8df3b4f6de7ae8bd1fc6a8bd1f31be56",
          "sha256": "c506387964fa1883d3b15f38988a2ceab740dc193e56aa53e0142824f4290997",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/tasksource/Boardgame-QA/78e38c3c8df3b4f6de7ae8bd1fc6a8bd1f31be56/README.md?%2Fdatasets%2Ftasksource%2FBoardgame-QA%2Fresolve%2F78e38c3c8df3b4f6de7ae8bd1fc6a8bd1f31be56%2FREADME.md=&etag=%2241653fdc1680483d6de026c99375779371c7872e%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "boardgameqa:default:test:0",
          "local_path": "output/typed_decisions/permissive-v1/cache/14443070e743849ef7e9ceb6d0ce1e5a372ceac3ba399ef2983ff546dda930b9",
          "original_row_match_status": "matched",
          "sha256": "8e10741e36463c8abcfdd29d93b4e55bf5360302144d0f6860ce05c4ed522e16",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://huggingface.co/datasets/tasksource/Boardgame-QA/resolve/866998af347330227cf19df823d3583ec8d77ec9/default/test/0000.parquet"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/permissive_adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "disproved": "The claim can be disproved.",
            "proved": "The claim can be proved.",
            "unknown": "Neither the claim nor its negation can be proved."
          },
          "id": "answer",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "Descriptions are copied from the pinned source conversion’s case criteria. Keep IDs aligned; only expand them from source definitions, never from the observed target.",
          "option_semantics": "fixed_source_taxonomy",
          "selector": {
            "decision_id": "answer",
            "type": "choice"
          }
        }
      ],
      "source_profile": "boardgameqa"
    },
    "canttalk": {
      "disposition": "include",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching.",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "b015ae8902d8429ce32ebfed9f9935f891d10ed5",
          "sha256": "56062a7d218e167f556bb1b7f09758dcfdda41ad14bfa7e3855002126b33a444",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/nvidia/CantTalkAboutThis-Topic-Control-Dataset/b015ae8902d8429ce32ebfed9f9935f891d10ed5/README.md?%2Fdatasets%2Fnvidia%2FCantTalkAboutThis-Topic-Control-Dataset%2Fresolve%2Fb015ae8902d8429ce32ebfed9f9935f891d10ed5%2FREADME.md=&etag=%221c90d7e006f3a17c33b0a7371c0b659574c2ea1b%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "canttalk:mixtral:test:0",
          "local_path": "output/typed_decisions/permissive-v1/cache/5a72a20f9b620aedcec3e0e9a32f92cb3219eabd154cd286dd568484ddec044b",
          "original_row_match_status": "matched",
          "sha256": "6871a9802fc3097f6605411946d6e66bcec6b923ed24bdb8d0e27c9460746010",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://huggingface.co/datasets/nvidia/CantTalkAboutThis-Topic-Control-Dataset/resolve/12765cfd2d479aa4f66d421df0b1b1b131e68978/mixtral/test/0000.parquet"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/permissive_adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "false": "The latest user message is outside the allowed topics in the supplied topic policy.",
            "true": "The latest user message stays within the allowed topics in the supplied topic policy."
          },
          "id": "answer",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "",
          "selector": {
            "decision_id": "answer",
            "type": "noul"
          }
        }
      ],
      "source_profile": "canttalk"
    },
    "civil_comments": {
      "disposition": "include_with_caveat",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching. Caveats: Existing source spot-check note: \"7属性の投票割合をNoulに保持。毒性の少数票や注釈判断の揺れも保持。属性の強度Scoreにしていない。\"",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "f2970eb3a55777454c94069077cc8d9b5866312d",
          "sha256": "8ce0fee0cce0d54ddd5daaad7ba7cd5ed31d78b8e8f2d5cd2345b0dd015932c6",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/google/civil_comments/f2970eb3a55777454c94069077cc8d9b5866312d/README.md?%2Fdatasets%2Fgoogle%2Fcivil_comments%2Fresolve%2Ff2970eb3a55777454c94069077cc8d9b5866312d%2FREADME.md=&etag=%2292511dd19fd36e335c9869acb33a35f1449d5a73%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "civil_comments:default:test:0",
          "local_path": "output/typed_decisions/v1/cache/b5e68fad318d6e83c825ef93043eac62a01ff670322192c825a66f7270504972",
          "original_row_match_status": "matched",
          "sha256": "410f803ff161e7c4ba005db7de5b444cac413f4861bf056b0fd69d670d75aa08",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://huggingface.co/datasets/google/civil_comments/resolve/29b3493cb776f0582ff4d6b37e40ef20ef0ce96d/default/test/0000.parquet"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "false": "The comment does not attack a person or group based on identity.",
            "true": "The comment attacks a person or group based on identity."
          },
          "id": "identity_attack",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "identity_attack",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The comment does not express an insult.",
            "true": "The comment expresses an insult."
          },
          "id": "insult",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "insult",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The comment does not contain obscene language.",
            "true": "The comment contains obscene language."
          },
          "id": "obscene",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "obscene",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The comment does not express severe toxicity.",
            "true": "The comment expresses severe toxicity."
          },
          "id": "severe_toxicity",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "severe_toxicity",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The comment does not contain sexually explicit content.",
            "true": "The comment contains sexually explicit content."
          },
          "id": "sexual_explicit",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "sexual_explicit",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The comment does not express a threat.",
            "true": "The comment expresses a threat."
          },
          "id": "threat",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "threat",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The comment is not rude, disrespectful, or unreasonable in a way likely to make people leave a discussion.",
            "true": "The comment is rude, disrespectful, or unreasonable in a way that is likely to make people leave a discussion."
          },
          "id": "toxicity",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "toxicity",
            "type": "noul"
          }
        }
      ],
      "source_profile": "civil_comments"
    },
    "cladder": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Original author archive at pinned commit; complete GitHub tree size manifest saved in source_manifests. No mirror license substitution.; No supervised training split exists in this release subset; treat available rows as evaluation-only/source-limited.",
      "evidence": [
        {
          "path_or_url": "https://github.com/causalNLP/cladder/blob/3d2d1169b4b939a09048a6a75956c8972a93cc38/README.md",
          "what_verified": "Upstream dataset/card identifier https://github.com/causalNLP/cladder at pinned revision 3d2d1169b4b939a09048a6a75956c8972a93cc38; license metadata is ['mit']. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://raw.githubusercontent.com/causalNLP/cladder/3d2d1169b4b939a09048a6a75956c8972a93cc38/README.md",
          "what_verified": "Primary source card was fetched and read at 3d2d1169b4b939a09048a6a75956c8972a93cc38; card excerpt: # CLadder: Assessing Causal Reasoning in Language Models ## 🚀 Get the dataset now! [cladder-v1.zip](https://github.com/causalNLP/cladder/raw/main/data/cladder-v1.zip)"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/cladder/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 1 source artifacts; 1 referenced local cache paths exist and integrity digests match. Plan source revision 3d2d1169b4b939a09048a6a75956c8972a93cc38."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/cladder/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/cladder",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 8917 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 3 actual rows from 1 pinned raw source files for source key cladder; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "false": "Under the causal assumptions stated in the input, the answer is no.",
            "true": "Under the causal assumptions stated in the input, the answer is yes."
          },
          "id": "cladder__noul__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Source-grounded false/true meanings; see source evidence.",
          "selector": {
            "decision_id": "answer",
            "kind": "judgment",
            "type": "noul"
          }
        }
      ],
      "source_profile": "cladder"
    },
    "clinc": {
      "disposition": "include_with_caveat",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching. Caveats: Existing source spot-check note: \"意図・候補説明の対応を確認。混合全体でMASSIVE評価発話と一致するtrain38件を除外し、再サンプルを確認。\"",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "155b9c710419136e17307b80d0a13e68cd46b4ec",
          "sha256": "e28704a3c04f6b05c286ca2a8891bbf32c69c000dedfc5825d6e1713369b7a89",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/clinc/clinc_oos/155b9c710419136e17307b80d0a13e68cd46b4ec/README.md?%2Fdatasets%2Fclinc%2Fclinc_oos%2Fresolve%2F155b9c710419136e17307b80d0a13e68cd46b4ec%2FREADME.md=&etag=%22676eddd998b36b7b132b4a4a6f5e27184a2824e2%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "clinc:plus:test:0",
          "local_path": "output/typed_decisions/v1/cache/5234426cdef890ff239e52b14a87f8adae4a0e43f8767293f79a49b73ada85f4",
          "original_row_match_status": "matched",
          "sha256": "3e60e45b25bf86543aa5df8ba4fcc674114164e6184f0197690648c2908d0102",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://huggingface.co/datasets/clinc/clinc_oos/resolve/8e951f061d8a5bedeaec359c02de3185e8c3359a/plus/test/0000.parquet"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "accept_reservations": "accept reservations",
            "account_blocked": "account blocked",
            "alarm": "alarm",
            "application_status": "application status",
            "apr": "apr",
            "are_you_a_bot": "are you a bot",
            "balance": "balance",
            "bill_balance": "bill balance",
            "bill_due": "bill due",
            "book_flight": "book flight",
            "book_hotel": "book hotel",
            "calculator": "calculator",
            "calendar": "calendar",
            "calendar_update": "calendar update",
            "calories": "calories",
            "cancel": "cancel",
            "cancel_reservation": "cancel reservation",
            "car_rental": "car rental",
            "card_declined": "card declined",
            "carry_on": "carry on",
            "change_accent": "change accent",
            "change_ai_name": "change ai name",
            "change_language": "change language",
            "change_speed": "change speed",
            "change_user_name": "change user name",
            "change_volume": "change volume",
            "confirm_reservation": "confirm reservation",
            "cook_time": "cook time",
            "credit_limit": "credit limit",
            "credit_limit_change": "credit limit change",
            "credit_score": "credit score",
            "current_location": "current location",
            "damaged_card": "damaged card",
            "date": "date",
            "definition": "definition",
            "direct_deposit": "direct deposit",
            "directions": "directions",
            "distance": "distance",
            "do_you_have_pets": "do you have pets",
            "exchange_rate": "exchange rate",
            "expiration_date": "expiration date",
            "find_phone": "find phone",
            "flight_status": "flight status",
            "flip_coin": "flip coin",
            "food_last": "food last",
            "freeze_account": "freeze account",
            "fun_fact": "fun fact",
            "gas": "gas",
            "gas_type": "gas type",
            "goodbye": "goodbye",
            "greeting": "greeting",
            "how_busy": "how busy",
            "how_old_are_you": "how old are you",
            "improve_credit_score": "improve credit score",
            "income": "income",
            "ingredient_substitution": "ingredient substitution",
            "ingredients_list": "ingredients list",
            "insurance": "insurance",
            "insurance_change": "insurance change",
            "interest_rate": "interest rate",
            "international_fees": "international fees",
            "international_visa": "international visa",
            "jump_start": "jump start",
            "last_maintenance": "last maintenance",
            "lost_luggage": "lost luggage",
            "make_call": "make call",
            "maybe": "maybe",
            "meal_suggestion": "meal suggestion",
            "meaning_of_life": "meaning of life",
            "measurement_conversion": "measurement conversion",
            "meeting_schedule": "meeting schedule",
            "min_payment": "min payment",
            "mpg": "mpg",
            "new_card": "new card",
            "next_holiday": "next holiday",
            "next_song": "next song",
            "no": "no",
            "nutrition_info": "nutrition info",
            "oil_change_how": "oil change how",
            "oil_change_when": "oil change when",
            "oos": "Outside all of the listed service intents.",
            "order": "order",
            "order_checks": "order checks",
            "order_status": "order status",
            "pay_bill": "pay bill",
            "payday": "payday",
            "pin_change": "pin change",
            "play_music": "play music",
            "plug_type": "plug type",
            "pto_balance": "pto balance",
            "pto_request": "pto request",
            "pto_request_status": "pto request status",
            "pto_used": "pto used",
            "recipe": "recipe",
            "redeem_rewards": "redeem rewards",
            "reminder": "reminder",
            "reminder_update": "reminder update",
            "repeat": "repeat",
            "replacement_card_duration": "replacement card duration",
            "report_fraud": "report fraud",
            "report_lost_card": "report lost card",
            "reset_settings": "reset settings",
            "restaurant_reservation": "restaurant reservation",
            "restaurant_reviews": "restaurant reviews",
            "restaurant_suggestion": "restaurant suggestion",
            "rewards_balance": "rewards balance",
            "roll_dice": "roll dice",
            "rollover_401k": "rollover 401k",
            "routing": "routing",
            "schedule_maintenance": "schedule maintenance",
            "schedule_meeting": "schedule meeting",
            "share_location": "share location",
            "shopping_list": "shopping list",
            "shopping_list_update": "shopping list update",
            "smart_home": "smart home",
            "spelling": "spelling",
            "spending_history": "spending history",
            "sync_device": "sync device",
            "taxes": "taxes",
            "tell_joke": "tell joke",
            "text": "text",
            "thank_you": "thank you",
            "time": "time",
            "timer": "timer",
            "timezone": "timezone",
            "tire_change": "tire change",
            "tire_pressure": "tire pressure",
            "todo_list": "todo list",
            "todo_list_update": "todo list update",
            "traffic": "traffic",
            "transactions": "transactions",
            "transfer": "transfer",
            "translate": "translate",
            "travel_alert": "travel alert",
            "travel_notification": "travel notification",
            "travel_suggestion": "travel suggestion",
            "uber": "uber",
            "update_playlist": "update playlist",
            "user_name": "user name",
            "vaccines": "vaccines",
            "w2": "w2",
            "weather": "weather",
            "what_are_your_hobbies": "what are your hobbies",
            "what_can_i_ask_you": "what can i ask you",
            "what_is_your_name": "what is your name",
            "what_song": "what song",
            "where_are_you_from": "where are you from",
            "whisper_mode": "whisper mode",
            "who_do_you_work_for": "who do you work for",
            "who_made_you": "who made you",
            "yes": "yes"
          },
          "id": "intent",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "Descriptions are copied from the pinned source conversion’s case criteria. Keep IDs aligned; only expand them from source definitions, never from the observed target.",
          "option_semantics": "fixed_source_taxonomy",
          "selector": {
            "decision_id": "intent",
            "type": "choice"
          }
        }
      ],
      "source_profile": "clinc"
    },
    "contract_nli": {
      "disposition": "include",
      "disposition_reason": "Original author archive at pinned commit; complete GitHub tree size manifest saved in source_manifests. No mirror license substitution.",
      "evidence": [
        {
          "path_or_url": "https://github.com/stanfordnlp/contract-nli/blob/eced6528dd3c1d14d73f9a87df8f7bdbc03126f9/README.md",
          "what_verified": "Upstream dataset/card identifier https://github.com/stanfordnlp/contract-nli at pinned revision eced6528dd3c1d14d73f9a87df8f7bdbc03126f9; license metadata is ['cc-by-4.0']. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://raw.githubusercontent.com/stanfordnlp/contract-nli/eced6528dd3c1d14d73f9a87df8f7bdbc03126f9/README.md",
          "what_verified": "Primary source card was fetched and read at eced6528dd3c1d14d73f9a87df8f7bdbc03126f9; card excerpt: # ContractNLI: A Dataset for Document-level Natural Language Inference for Contracts Please visit the website at https://stanfordnlp.github.io/contract-nli/ ."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/contract_nli/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 1 source artifacts; 1 referenced local cache paths exist and integrity digests match. Plan source revision eced6528dd3c1d14d73f9a87df8f7bdbc03126f9."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/contract_nli/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/contract_nli",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 606 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 1 actual rows from 1 pinned raw source files for source key contract_nli; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "id": "contract_nli__choice__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Preserve source Choice option IDs/descriptions and order; sampled option lists can be instance-specific.",
          "selector": {
            "decision_id": "nda\\-\\d+",
            "kind": "judgment",
            "type": "choice"
          }
        }
      ],
      "source_profile": "contract_nli"
    },
    "corr2cause": {
      "disposition": "include",
      "disposition_reason": "See permissive dataset build note; short task-specific instructions, source-specific label checks and grouped splits. Author license overrides missing mirror metadata where documented.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/tasksource/corr2cause/blob/aa0e8b909b8cb7589eb23dad5e9c1cb709e665a3/README.md",
          "what_verified": "Upstream dataset/card identifier tasksource/corr2cause at pinned revision aa0e8b909b8cb7589eb23dad5e9c1cb709e665a3; license metadata is ['mit']. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/tasksource/corr2cause/raw/aa0e8b909b8cb7589eb23dad5e9c1cb709e665a3/README.md",
          "what_verified": "Primary source card was fetched and read at aa0e8b909b8cb7589eb23dad5e9c1cb709e665a3; card excerpt: --- license: mit dataset_info: features: - name: premise dtype: string - name: hypothesis dtype: string"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/corr2cause/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 4 source artifacts; 4 referenced local cache paths exist and integrity digests match. Plan source revision aa0e8b909b8cb7589eb23dad5e9c1cb709e665a3."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/corr2cause/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/corr2cause",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 277084 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 9 actual rows from 3 pinned raw source files for source key corr2cause; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "id": "corr2cause__choice__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Preserve source Choice option IDs/descriptions and order; sampled option lists can be instance-specific.",
          "selector": {
            "decision_id": "relation",
            "kind": "judgment",
            "type": "choice"
          }
        }
      ],
      "source_profile": "corr2cause"
    },
    "creak": {
      "disposition": "include",
      "disposition_reason": "Reviewed English typed judgment conversion; preserve selected source splits and label semantics.; An unlabeled upstream split is present; keep it unlabeled and out of supervised target loss.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/amydeng2000/CREAK/blob/cceb4696560317e920d6512b906263bb425883a1/README.md",
          "what_verified": "Upstream dataset/card identifier amydeng2000/CREAK at pinned revision cceb4696560317e920d6512b906263bb425883a1; license metadata is ['mit']. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/amydeng2000/CREAK/raw/cceb4696560317e920d6512b906263bb425883a1/README.md",
          "what_verified": "Primary source card was fetched and read at cceb4696560317e920d6512b906263bb425883a1; card excerpt: Home page & Original source: https://github.com/yasumasaonoe/creak"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/creak/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 4 source artifacts; 4 referenced local cache paths exist and integrity digests match. Plan source revision cceb4696560317e920d6512b906263bb425883a1."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/creak/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/creak",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 13383 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 12 actual rows from 4 pinned raw source files for source key creak; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "false": "The claim is factually false.",
            "true": "The claim is factually true."
          },
          "id": "creak__noul__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Source-grounded false/true meanings; see source evidence.",
          "selector": {
            "decision_id": "true",
            "kind": "judgment",
            "type": "noul"
          }
        }
      ],
      "source_profile": "creak"
    },
    "crows_pairs": {
      "disposition": "include",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching.",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "8aaac11c485473159ec9328a65253a5be9a479dc",
          "sha256": "bf96decb803c46ae0cf5ae065ac374ccead2bff74c62a258621ffbed5db1925b",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://raw.githubusercontent.com/nyu-mll/crows-pairs/8aaac11c485473159ec9328a65253a5be9a479dc/README.md"
        },
        {
          "adapter_replay": "matched",
          "case_id": "crows_pairs:default:test:0",
          "local_path": "output/typed_decisions/additional20-v1/cache/e5bdb81270a3eacb2017c549379974cdbb45e3088c6337686ef703e1b317fda0",
          "original_row_match_status": "matched",
          "sha256": "dfb36986ce0502abbaf7055b9176da3d08d48e07df1251991b5dfbcbceab9d0c",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://raw.githubusercontent.com/nyu-mll/crows-pairs/8aaac11c485473159ec9328a65253a5be9a479dc/data/crows_pairs_anonymized.csv"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/additional_adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {},
          "id": "stereotype",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "Options and their descriptions vary by case; use the row’s own option-ID/description pairs. Do not assign a fixed meaning to option IDs across rows.",
          "option_semantics": "per_case_candidates",
          "preserve_case_options": true,
          "selector": {
            "decision_id": "stereotype",
            "type": "choice"
          }
        }
      ],
      "source_profile": "crows_pairs"
    },
    "dbpedia": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Reviewed English typed judgment conversion; preserve selected source splits and label semantics.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/fancyzhx/dbpedia_14/blob/9abd46cf7fc8b4c64290f26993c540b92aa145ac/README.md",
          "what_verified": "Upstream dataset/card identifier fancyzhx/dbpedia_14 at pinned revision 9abd46cf7fc8b4c64290f26993c540b92aa145ac; license metadata is ['cc-by-sa-3.0']. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/fancyzhx/dbpedia_14/raw/9abd46cf7fc8b4c64290f26993c540b92aa145ac/README.md",
          "what_verified": "Primary source card was fetched and read at 9abd46cf7fc8b4c64290f26993c540b92aa145ac; card excerpt: # Dataset Card for DBpedia14 ## Table of Contents - [Dataset Description](#dataset-description) - [Dataset Summary](#dataset-summary)"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/dbpedia/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 2 source artifacts; 2 referenced local cache paths exist and integrity digests match. Plan source revision 9abd46cf7fc8b4c64290f26993c540b92aa145ac."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/dbpedia/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/dbpedia",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 629999 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 6 actual rows from 2 pinned raw source files for source key dbpedia; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "id": "dbpedia__choice__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Preserve source Choice option IDs/descriptions and order; sampled option lists can be instance-specific.",
          "selector": {
            "decision_id": "label",
            "kind": "judgment",
            "type": "choice"
          }
        }
      ],
      "source_profile": "dbpedia"
    },
    "defeasible_nli": {
      "disposition": "include_with_caveat",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching. Caveats: Existing source spot-check note: \"updateが仮説を強めるか弱めるかを学習し、仮説そのものの真偽とは区別。直観に合わないweakener例も元注釈のまま。SNLI由来のみ採用。\"",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "7c4a57df9d8de5c36d4e9caa977907b5e8469c4f",
          "sha256": "13d1634dc80dcdd7080fda08f13fa18f2b7401e4034d0ec17cedd0683b830d13",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/tasksource/defeasible-nli/7c4a57df9d8de5c36d4e9caa977907b5e8469c4f/README.md?%2Fdatasets%2Ftasksource%2Fdefeasible-nli%2Fresolve%2F7c4a57df9d8de5c36d4e9caa977907b5e8469c4f%2FREADME.md=&etag=%22f39a80c6e989bb5200b69244e87edb654288b20e%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "defeasible_nli:snli:test:0",
          "local_path": "output/typed_decisions/v1/cache/2dc516af0cb7d0e1526fc271e502efb9c17d476ffb802b51a01c33bbcb6ae29f",
          "original_row_match_status": "matched",
          "sha256": "9f0c0f413d8be45fd88c92af58993164e0b499a1a1e73ae394ba1c3868850607",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://huggingface.co/datasets/tasksource/defeasible-nli/resolve/6fe21a0a12a0ff81b818288904451970adf20edc/snli/test/0000.parquet"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/extra_adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "strengthener": "The update makes the hypothesis more likely given the premise.",
            "weakener": "The update makes the hypothesis less likely given the premise."
          },
          "id": "update_effect",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "Descriptions are copied from the pinned source conversion’s case criteria. Keep IDs aligned; only expand them from source definitions, never from the observed target.",
          "option_semantics": "fixed_source_taxonomy",
          "selector": {
            "decision_id": "update_effect",
            "type": "choice"
          }
        }
      ],
      "source_profile": "defeasible_nli"
    },
    "laya__enron_spam": {
      "disposition": "include",
      "disposition_reason": "Laya base training source; source card does not specify a dataset license.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/SetFit/enron_spam/blob/main/README.md",
          "what_verified": "Upstream dataset/card identifier SetFit/enron_spam at pinned revision 1916f66c89d52221ae33eb57d44498b4f3a5df22; license metadata is unknown. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/SetFit/enron_spam/raw/1916f66c89d52221ae33eb57d44498b4f3a5df22/README.md",
          "what_verified": "Primary source card was fetched and read at 1916f66c89d52221ae33eb57d44498b4f3a5df22; card excerpt: This is a version of the [Enron Spam Email Dataset](https://github.com/MWiechmann/enron_spam_data), containing emails (subject + message) and a label whether it is spam or ham."
        },
        {
          "path_or_url": "output/datasets/bekko-system-one-traing-data-v1/laya/enron_spam/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 2 source artifacts; 2 referenced local cache paths exist and integrity digests match. Plan source revision 1916f66c89d52221ae33eb57d44498b4f3a5df22."
        },
        {
          "path_or_url": "output/datasets/bekko-system-one-traing-data-v1/laya/enron_spam/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/datasets/bekko-system-one-traing-data-v1/laya/enron_spam",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 30101 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/laya_adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 6 actual rows from 2 pinned raw source files for source key enron_spam; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "false": "The email is not unsolicited spam or bulk marketing.",
            "true": "The email is unsolicited spam or bulk marketing."
          },
          "id": "enron_spam__noul__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Source-grounded false/true meanings; see source evidence.",
          "selector": {
            "decision_id": "is_spam",
            "kind": "judgment",
            "type": "noul"
          }
        }
      ],
      "source_profile": "laya__enron_spam"
    },
    "esci": {
      "disposition": "include",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching.",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "8113b17a5d4099e20243282c926f1bc1a08a4d13",
          "sha256": "0f40b428b589178f461b42f44ddf7d19eb4146277fc3e815ec4534b629dd4103",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/tasksource/esci/8113b17a5d4099e20243282c926f1bc1a08a4d13/README.md?%2Fdatasets%2Ftasksource%2Fesci%2Fresolve%2F8113b17a5d4099e20243282c926f1bc1a08a4d13%2FREADME.md=&etag=%22d80a8be22acade64071091678abff1c11076c83e%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "esci:default:test:1",
          "local_path": "output/typed_decisions/permissive-v1/cache/9ee684572c5b90f5304ace11307dda5dff5ddf4024382da6cff34a959d558ccd",
          "original_row_match_status": "matched",
          "sha256": "01b0da5ffc4d8e414ae1d04e98c89bd06ebb35196796de1fd99032a775b04a8e",
          "source": "pinned_raw_example",
          "source_row_index": 1,
          "state_and_decisions_matched": true,
          "url": "https://huggingface.co/datasets/tasksource/esci/resolve/45c948250c2116f1e535bac67b92501c695307a4/default/test/0000.parquet"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/permissive_adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "Complement": "A complementary product, not a substitute.",
            "Exact": "Matches the requested product.",
            "Irrelevant": "Does not meet or complement the request.",
            "Substitute": "A usable alternative to the requested product."
          },
          "id": "relevance",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "Descriptions are copied from the pinned source conversion’s case criteria. Keep IDs aligned; only expand them from source definitions, never from the observed target.",
          "option_semantics": "fixed_source_taxonomy",
          "selector": {
            "decision_id": "relevance",
            "type": "choice"
          }
        }
      ],
      "source_profile": "esci"
    },
    "ethics": {
      "disposition": "include",
      "disposition_reason": "See research note; adapters and full-data quality checks are not implemented by this survey.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/hendrycks/ethics/blob/b8b47c589f8bee77175b8648e5497278b68da48a/README.md",
          "what_verified": "Upstream dataset/card identifier hendrycks/ethics at pinned revision b8b47c589f8bee77175b8648e5497278b68da48a; license metadata is ['mit']. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/hendrycks/ethics/raw/b8b47c589f8bee77175b8648e5497278b68da48a/README.md",
          "what_verified": "Primary source card was fetched and read at b8b47c589f8bee77175b8648e5497278b68da48a; card excerpt: # Dataset Card for ETHICS This is the data from [Aligning AI With Shared Human Values](https://arxiv.org/pdf/2008.02275) by Dan Hendrycks, Collin Burns, Steven Basart, Andrew Critch,  Jerry Li, Dawn Song, and Jacob Steinhardt, published at ICLR 2021. For more information, see the [Github Repo](https://github.com/hendrycks/ethics)."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/ethics/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 15 source artifacts; 15 referenced local cache paths exist and integrity digests match. Plan source revision b8b47c589f8bee77175b8648e5497278b68da48a."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/ethics/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/ethics",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 125552 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 45 actual rows from 15 pinned raw source files for source key ethics; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "false": "The described action is not morally wrong.",
            "true": "The described action is morally wrong."
          },
          "id": "ethics__noul__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Source-grounded false/true meanings; see source evidence.",
          "selector": {
            "decision_id": "commonsense",
            "kind": "judgment",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The excuse or obligation is not reasonable in the scenario.",
            "true": "The excuse or obligation is reasonable in the scenario."
          },
          "id": "ethics__noul__001",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Source-grounded false/true meanings; see source evidence.",
          "selector": {
            "decision_id": "deontology",
            "kind": "judgment",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The described treatment is not fair given its stated reason.",
            "true": "The described treatment is fair given its stated reason."
          },
          "id": "ethics__noul__002",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Source-grounded false/true meanings; see source evidence.",
          "selector": {
            "decision_id": "justice",
            "kind": "judgment",
            "type": "noul"
          }
        },
        {
          "id": "ethics__choice__003",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Preserve source Choice option IDs/descriptions and order; sampled option lists can be instance-specific.",
          "selector": {
            "decision_id": "pleasantness",
            "kind": "judgment",
            "type": "choice"
          }
        },
        {
          "criteria_by_id": {
            "false": "The trait in the task instruction does not apply to the described person.",
            "true": "The trait in the task instruction applies to the described person."
          },
          "id": "ethics__noul__004",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Source-grounded false/true meanings; see source evidence.",
          "selector": {
            "decision_id": "virtue",
            "kind": "judgment",
            "type": "noul"
          }
        }
      ],
      "source_profile": "ethics"
    },
    "ethos": {
      "disposition": "include_with_caveat",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching. Caveats: Existing source spot-check note: \"identityへのhateと一般的な侮辱の区別を確認。元テキストに文字化けあり。採用splitはSetFitミラーの既存splitであり原著者の公式splitとは断定しない。\"",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "3e20849175072aa95845981f27938ebeb4eb93e0",
          "sha256": "fef6b3abafb515355a56ad82f02ebaf8d7b491ad29f59a6045eff9baeb169328",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/SetFit/ethos_binary/3e20849175072aa95845981f27938ebeb4eb93e0/README.md?%2Fdatasets%2FSetFit%2Fethos_binary%2Fresolve%2F3e20849175072aa95845981f27938ebeb4eb93e0%2FREADME.md=&etag=%22718a0d604f139a77b3536a36cfe8e0a82d385be0%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "ethos:default:test:0",
          "local_path": "output/typed_decisions/v1/cache/aff7f595782d64be65c1c17d5b4a7557cf1763af0e12e0fc2f21397194ae1e7a",
          "original_row_match_status": "matched",
          "sha256": "ae7080e00a3668ffae38ca4fef3728c8a43fd8374e44cf69eed7c4dd2ab6451f",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://huggingface.co/datasets/SetFit/ethos_binary/resolve/b6a170577d6745c96e26a4c51a18342f5b76dd6b/default/test/0000.parquet"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/extra_adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "false": "The text does not express identity-directed hate; a general insult by itself is not enough.",
            "true": "The text expresses hate toward a person or group based on identity."
          },
          "id": "hateful",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "",
          "selector": {
            "decision_id": "hateful",
            "type": "noul"
          }
        }
      ],
      "source_profile": "ethos"
    },
    "few_nerd": {
      "disposition": "include_with_caveat",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching. Caveats: Existing source spot-check note: \"既知のspanを8粗カテゴリへ分類。抽出タスクではない。validationのRaytheon=person、Raytheon Aircraft=productは元Parquetのタグでも同じで、変換の取り違えではないが注釈誤りが疑われる。BIO境界がないため同一fine tagの連続区間をspanとする限界がある。\"",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "205f3e9c9f3577ea2561d43f2f62dc249ab92d5b",
          "sha256": "beab6f6b7887bc882b88ca7691f6e83316edde29d07fea5491a35a6ffb0cad74",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/DFKI-SLT/few-nerd/205f3e9c9f3577ea2561d43f2f62dc249ab92d5b/README.md?%2Fdatasets%2FDFKI-SLT%2Ffew-nerd%2Fresolve%2F205f3e9c9f3577ea2561d43f2f62dc249ab92d5b%2FREADME.md=&etag=%22e25034e6952958ffb5587f106b72603efd244dd3%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "few_nerd:supervised:test:0",
          "local_path": "output/typed_decisions/v1/cache/35929ae7d68b13ee4cf83cd88aede862c5081c416db57acea91028474f14154c",
          "original_row_match_status": "matched",
          "sha256": "b7ad746fcbeb68fcc235ba7142d7c3723ea2dc39930089e947284defecf300c6",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://huggingface.co/datasets/DFKI-SLT/few-nerd/resolve/0ed916112325e1e36b49b9528dfad44835f641b8/supervised/test/0000.parquet"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/extra_adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "art": "art",
            "building": "building",
            "event": "event",
            "location": "location",
            "organization": "organization",
            "other": "other",
            "person": "person",
            "product": "product"
          },
          "id": "entity_span_pattern",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "Descriptions are copied from the pinned source conversion’s case criteria. Keep IDs aligned; only expand them from source definitions, never from the observed target.",
          "option_semantics": "fixed_source_taxonomy",
          "selector": {
            "decision_id": "^entity_[0-9]+_[0-9]+$",
            "type": "choice"
          }
        }
      ],
      "source_profile": "few_nerd"
    },
    "fol_nli": {
      "disposition": "include",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching.",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "6b8a2ec01b226ed871fbde4677a31146ac5f370c",
          "sha256": "41d29653780e58f1ad45acb95d30ebe5bc837acbe3e1f6451aedc206b1291a16",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/tasksource/FOL-nli/6b8a2ec01b226ed871fbde4677a31146ac5f370c/README.md?%2Fdatasets%2Ftasksource%2FFOL-nli%2Fresolve%2F6b8a2ec01b226ed871fbde4677a31146ac5f370c%2FREADME.md=&etag=%22e9c34b59425eb67b741cbd1bf0c735548d42b98c%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "fol_nli:default:test:0",
          "local_path": "output/typed_decisions/permissive-v1/cache/08b2338f75ebe221ccb42a63d2124b32fa0c9e93a2f0d989608ec3332f199bf3",
          "original_row_match_status": "matched",
          "sha256": "9f5df7f27bca723d275a5f4f44a577726f820705121db164cd5a34ff41e06469",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://huggingface.co/datasets/tasksource/FOL-nli/resolve/345040911479f3b00a3f9763b0b01b90b6d0b2b8/default/test/0000.parquet"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/permissive_adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "contradiction": "The premise guarantees that the hypothesis is false.",
            "entailment": "The premise guarantees that the hypothesis is true.",
            "neutral": "The premise does not determine whether the hypothesis is true."
          },
          "id": "relation",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "Descriptions are copied from the pinned source conversion’s case criteria. Keep IDs aligned; only expand them from source definitions, never from the observed target.",
          "option_semantics": "fixed_source_taxonomy",
          "selector": {
            "decision_id": "relation",
            "type": "choice"
          }
        }
      ],
      "source_profile": "fol_nli"
    },
    "followir_core17": {
      "disposition": "include_with_caveat",
      "disposition_reason": "No local source plan/report was present alongside the export; upstream revision is metadata-only.; Training judgments are synthetic (source_json annotation states GPT-3.5 synthetic document/label; Mistral-filtered). Preserve this provenance. Test decisions pair original/changed instructions within the same case; keep pairs together and evaluate instruction-following effects. This derived Noul evaluation is not official p-MRR.; No supervised training split exists in this release subset; treat available rows as evaluation-only/source-limited.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/mteb/Core17InstructionRetrieval/blob/67d8733dd0691f08f34ed9dda5941578cba8d91c/README.md",
          "what_verified": "Upstream dataset/card identifier mteb/Core17InstructionRetrieval at pinned revision 67d8733dd0691f08f34ed9dda5941578cba8d91c; license metadata is mit. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/mteb/News21InstructionRetrieval/blob/c6fffeb9cdd95c1ffa81c62c687b48b56352872b/README.md",
          "what_verified": "Upstream dataset/card identifier mteb/News21InstructionRetrieval at pinned revision c6fffeb9cdd95c1ffa81c62c687b48b56352872b; license metadata is mit. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/mteb/Robust04InstructionRetrieval/blob/0495a5cb69aa8fa5bca81bd56ac248911c19eeb6/README.md",
          "what_verified": "Upstream dataset/card identifier mteb/Robust04InstructionRetrieval at pinned revision 0495a5cb69aa8fa5bca81bd56ac248911c19eeb6; license metadata is mit. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/jhu-clsp/FollowIR-train/blob/545192353a0192e2ed0e0e4f1604ff8a231f8733/README.md",
          "what_verified": "Upstream dataset/card identifier jhu-clsp/FollowIR-train at pinned revision 545192353a0192e2ed0e0e4f1604ff8a231f8733; license metadata is apache-2.0. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/mteb/Core17InstructionRetrieval/raw/67d8733dd0691f08f34ed9dda5941578cba8d91c/README.md",
          "what_verified": "Primary source card was fetched and read at 67d8733dd0691f08f34ed9dda5941578cba8d91c; card excerpt: ## How to evaluate on this task You can evaluate an embedding model on this dataset using the following code: ```python"
        },
        {
          "path_or_url": "https://huggingface.co/datasets/mteb/News21InstructionRetrieval/raw/c6fffeb9cdd95c1ffa81c62c687b48b56352872b/README.md",
          "what_verified": "Primary source card was fetched and read at c6fffeb9cdd95c1ffa81c62c687b48b56352872b; card excerpt: ## How to evaluate on this task You can evaluate an embedding model on this dataset using the following code: ```python"
        },
        {
          "path_or_url": "https://huggingface.co/datasets/mteb/Robust04InstructionRetrieval/raw/0495a5cb69aa8fa5bca81bd56ac248911c19eeb6/README.md",
          "what_verified": "Primary source card was fetched and read at 0495a5cb69aa8fa5bca81bd56ac248911c19eeb6; card excerpt: ## How to evaluate on this task You can evaluate an embedding model on this dataset using the following code: ```python"
        },
        {
          "path_or_url": "https://huggingface.co/datasets/jhu-clsp/FollowIR-train/raw/545192353a0192e2ed0e0e4f1604ff8a231f8733/README.md",
          "what_verified": "Primary source card was fetched and read at 545192353a0192e2ed0e0e4f1604ff8a231f8733; card excerpt: # Dataset Summary FollowIR-train contains ~1800 query and instruction pairs, with labels for relevance (true or false). It can be used to train retrieval models to better follow instructions (see [FollowIR-7B](https://huggingface.co/jhu-clsp/FollowIR-7B)). The dataset was created by taking instruction and query pairs from all [TREC tracks](https://trec.nist.gov/) (which provides instructions as \"narratives\") from 1993-on that provided these instructions. Synthetic documents were then created fro"
        },
        {
          "path_or_url": "output/typed_decisions/followir-v1/release/upstream-source-source-plan.json",
          "what_verified": "No source plan file found."
        },
        {
          "path_or_url": "output/typed_decisions/followir-v1/release/upstream-source-report.json",
          "what_verified": "No upstream-source report found."
        },
        {
          "path_or_url": "output/typed_decisions/followir-v1/release",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has ? source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 54 actual rows from 18 pinned raw source files for source key followir; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/research_dataset/followir-source-v1",
          "what_verified": "Inspected FollowIR train/core17/news21/robust04 source plans and raw cache artifacts; plan URLs matched cache sidecars and checksums. Dedicated test components use separate task instructions and must retain paired/source-specific instructions."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "false": "The document fails at least one relevance requirement in the supplied instruction.",
            "true": "The document satisfies the relevance requirements in the supplied instruction."
          },
          "id": "followir__noul__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Source-grounded false/true meanings; see source evidence.",
          "selector": {
            "decision_id": "relevance_\\d+",
            "kind": "judgment",
            "type": "noul"
          }
        }
      ],
      "source_profile": "followir"
    },
    "followir_news21": {
      "disposition": "include_with_caveat",
      "disposition_reason": "No local source plan/report was present alongside the export; upstream revision is metadata-only.; Training judgments are synthetic (source_json annotation states GPT-3.5 synthetic document/label; Mistral-filtered). Preserve this provenance. Test decisions pair original/changed instructions within the same case; keep pairs together and evaluate instruction-following effects. This derived Noul evaluation is not official p-MRR.; No supervised training split exists in this release subset; treat available rows as evaluation-only/source-limited.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/mteb/Core17InstructionRetrieval/blob/67d8733dd0691f08f34ed9dda5941578cba8d91c/README.md",
          "what_verified": "Upstream dataset/card identifier mteb/Core17InstructionRetrieval at pinned revision 67d8733dd0691f08f34ed9dda5941578cba8d91c; license metadata is mit. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/mteb/News21InstructionRetrieval/blob/c6fffeb9cdd95c1ffa81c62c687b48b56352872b/README.md",
          "what_verified": "Upstream dataset/card identifier mteb/News21InstructionRetrieval at pinned revision c6fffeb9cdd95c1ffa81c62c687b48b56352872b; license metadata is mit. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/mteb/Robust04InstructionRetrieval/blob/0495a5cb69aa8fa5bca81bd56ac248911c19eeb6/README.md",
          "what_verified": "Upstream dataset/card identifier mteb/Robust04InstructionRetrieval at pinned revision 0495a5cb69aa8fa5bca81bd56ac248911c19eeb6; license metadata is mit. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/jhu-clsp/FollowIR-train/blob/545192353a0192e2ed0e0e4f1604ff8a231f8733/README.md",
          "what_verified": "Upstream dataset/card identifier jhu-clsp/FollowIR-train at pinned revision 545192353a0192e2ed0e0e4f1604ff8a231f8733; license metadata is apache-2.0. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/mteb/Core17InstructionRetrieval/raw/67d8733dd0691f08f34ed9dda5941578cba8d91c/README.md",
          "what_verified": "Primary source card was fetched and read at 67d8733dd0691f08f34ed9dda5941578cba8d91c; card excerpt: ## How to evaluate on this task You can evaluate an embedding model on this dataset using the following code: ```python"
        },
        {
          "path_or_url": "https://huggingface.co/datasets/mteb/News21InstructionRetrieval/raw/c6fffeb9cdd95c1ffa81c62c687b48b56352872b/README.md",
          "what_verified": "Primary source card was fetched and read at c6fffeb9cdd95c1ffa81c62c687b48b56352872b; card excerpt: ## How to evaluate on this task You can evaluate an embedding model on this dataset using the following code: ```python"
        },
        {
          "path_or_url": "https://huggingface.co/datasets/mteb/Robust04InstructionRetrieval/raw/0495a5cb69aa8fa5bca81bd56ac248911c19eeb6/README.md",
          "what_verified": "Primary source card was fetched and read at 0495a5cb69aa8fa5bca81bd56ac248911c19eeb6; card excerpt: ## How to evaluate on this task You can evaluate an embedding model on this dataset using the following code: ```python"
        },
        {
          "path_or_url": "https://huggingface.co/datasets/jhu-clsp/FollowIR-train/raw/545192353a0192e2ed0e0e4f1604ff8a231f8733/README.md",
          "what_verified": "Primary source card was fetched and read at 545192353a0192e2ed0e0e4f1604ff8a231f8733; card excerpt: # Dataset Summary FollowIR-train contains ~1800 query and instruction pairs, with labels for relevance (true or false). It can be used to train retrieval models to better follow instructions (see [FollowIR-7B](https://huggingface.co/jhu-clsp/FollowIR-7B)). The dataset was created by taking instruction and query pairs from all [TREC tracks](https://trec.nist.gov/) (which provides instructions as \"narratives\") from 1993-on that provided these instructions. Synthetic documents were then created fro"
        },
        {
          "path_or_url": "output/typed_decisions/followir-v1/release/upstream-source-source-plan.json",
          "what_verified": "No source plan file found."
        },
        {
          "path_or_url": "output/typed_decisions/followir-v1/release/upstream-source-report.json",
          "what_verified": "No upstream-source report found."
        },
        {
          "path_or_url": "output/typed_decisions/followir-v1/release",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has ? source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 54 actual rows from 18 pinned raw source files for source key followir; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/research_dataset/followir-source-v1",
          "what_verified": "Inspected FollowIR train/core17/news21/robust04 source plans and raw cache artifacts; plan URLs matched cache sidecars and checksums. Dedicated test components use separate task instructions and must retain paired/source-specific instructions."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "false": "The document fails at least one relevance requirement in the supplied instruction.",
            "true": "The document satisfies the relevance requirements in the supplied instruction."
          },
          "id": "followir__noul__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Source-grounded false/true meanings; see source evidence.",
          "selector": {
            "decision_id": "relevance_\\d+",
            "kind": "judgment",
            "type": "noul"
          }
        }
      ],
      "source_profile": "followir"
    },
    "followir_robust04": {
      "disposition": "include_with_caveat",
      "disposition_reason": "No local source plan/report was present alongside the export; upstream revision is metadata-only.; Training judgments are synthetic (source_json annotation states GPT-3.5 synthetic document/label; Mistral-filtered). Preserve this provenance. Test decisions pair original/changed instructions within the same case; keep pairs together and evaluate instruction-following effects. This derived Noul evaluation is not official p-MRR.; No supervised training split exists in this release subset; treat available rows as evaluation-only/source-limited.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/mteb/Core17InstructionRetrieval/blob/67d8733dd0691f08f34ed9dda5941578cba8d91c/README.md",
          "what_verified": "Upstream dataset/card identifier mteb/Core17InstructionRetrieval at pinned revision 67d8733dd0691f08f34ed9dda5941578cba8d91c; license metadata is mit. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/mteb/News21InstructionRetrieval/blob/c6fffeb9cdd95c1ffa81c62c687b48b56352872b/README.md",
          "what_verified": "Upstream dataset/card identifier mteb/News21InstructionRetrieval at pinned revision c6fffeb9cdd95c1ffa81c62c687b48b56352872b; license metadata is mit. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/mteb/Robust04InstructionRetrieval/blob/0495a5cb69aa8fa5bca81bd56ac248911c19eeb6/README.md",
          "what_verified": "Upstream dataset/card identifier mteb/Robust04InstructionRetrieval at pinned revision 0495a5cb69aa8fa5bca81bd56ac248911c19eeb6; license metadata is mit. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/jhu-clsp/FollowIR-train/blob/545192353a0192e2ed0e0e4f1604ff8a231f8733/README.md",
          "what_verified": "Upstream dataset/card identifier jhu-clsp/FollowIR-train at pinned revision 545192353a0192e2ed0e0e4f1604ff8a231f8733; license metadata is apache-2.0. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/mteb/Core17InstructionRetrieval/raw/67d8733dd0691f08f34ed9dda5941578cba8d91c/README.md",
          "what_verified": "Primary source card was fetched and read at 67d8733dd0691f08f34ed9dda5941578cba8d91c; card excerpt: ## How to evaluate on this task You can evaluate an embedding model on this dataset using the following code: ```python"
        },
        {
          "path_or_url": "https://huggingface.co/datasets/mteb/News21InstructionRetrieval/raw/c6fffeb9cdd95c1ffa81c62c687b48b56352872b/README.md",
          "what_verified": "Primary source card was fetched and read at c6fffeb9cdd95c1ffa81c62c687b48b56352872b; card excerpt: ## How to evaluate on this task You can evaluate an embedding model on this dataset using the following code: ```python"
        },
        {
          "path_or_url": "https://huggingface.co/datasets/mteb/Robust04InstructionRetrieval/raw/0495a5cb69aa8fa5bca81bd56ac248911c19eeb6/README.md",
          "what_verified": "Primary source card was fetched and read at 0495a5cb69aa8fa5bca81bd56ac248911c19eeb6; card excerpt: ## How to evaluate on this task You can evaluate an embedding model on this dataset using the following code: ```python"
        },
        {
          "path_or_url": "https://huggingface.co/datasets/jhu-clsp/FollowIR-train/raw/545192353a0192e2ed0e0e4f1604ff8a231f8733/README.md",
          "what_verified": "Primary source card was fetched and read at 545192353a0192e2ed0e0e4f1604ff8a231f8733; card excerpt: # Dataset Summary FollowIR-train contains ~1800 query and instruction pairs, with labels for relevance (true or false). It can be used to train retrieval models to better follow instructions (see [FollowIR-7B](https://huggingface.co/jhu-clsp/FollowIR-7B)). The dataset was created by taking instruction and query pairs from all [TREC tracks](https://trec.nist.gov/) (which provides instructions as \"narratives\") from 1993-on that provided these instructions. Synthetic documents were then created fro"
        },
        {
          "path_or_url": "output/typed_decisions/followir-v1/release/upstream-source-source-plan.json",
          "what_verified": "No source plan file found."
        },
        {
          "path_or_url": "output/typed_decisions/followir-v1/release/upstream-source-report.json",
          "what_verified": "No upstream-source report found."
        },
        {
          "path_or_url": "output/typed_decisions/followir-v1/release",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has ? source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 54 actual rows from 18 pinned raw source files for source key followir; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/research_dataset/followir-source-v1",
          "what_verified": "Inspected FollowIR train/core17/news21/robust04 source plans and raw cache artifacts; plan URLs matched cache sidecars and checksums. Dedicated test components use separate task instructions and must retain paired/source-specific instructions."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "false": "The document fails at least one relevance requirement in the supplied instruction.",
            "true": "The document satisfies the relevance requirements in the supplied instruction."
          },
          "id": "followir__noul__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Source-grounded false/true meanings; see source evidence.",
          "selector": {
            "decision_id": "relevance_\\d+",
            "kind": "judgment",
            "type": "noul"
          }
        }
      ],
      "source_profile": "followir"
    },
    "go_emotions": {
      "disposition": "include_with_caveat",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching. Caveats: Existing source spot-check note: \"27感情+neutralを独立Noulとして有効票を保持。同一文で複数感情がtrueでも正しい。rawの公式test混入を防止。\"",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "add492243ff905527e67aeb8b80c082af02207c3",
          "sha256": "0a5a9f3a949cce822f6d7453b91433b9fac739187f51ecc665fc5b8a6ec291ba",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/google-research-datasets/go_emotions/add492243ff905527e67aeb8b80c082af02207c3/README.md?%2Fdatasets%2Fgoogle-research-datasets%2Fgo_emotions%2Fresolve%2Fadd492243ff905527e67aeb8b80c082af02207c3%2FREADME.md=&etag=%228f6f3e58175acc3822053ef374724816605dc79d%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "go_emotions:raw_votes_official_split:test:0",
          "local_path": "output/typed_decisions/v1/cache/f0cffaf3a4b0417f74af835fbb203545aa59e1d5b5b1f66ca509caaeda52af3f",
          "original_row_match_status": "matched",
          "sha256": "fd0953e535ba2569edc6a1daaa1133f8e4b9071691d540c9bab812fda132bc26",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://huggingface.co/datasets/google-research-datasets/go_emotions/resolve/8bc92813f1fe1e029bbef32f58c507350111f851/simplified/test/0000.parquet"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "false": "The text does not express admiration.",
            "true": "The text expresses admiration."
          },
          "id": "admiration",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "admiration",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The text does not express amusement.",
            "true": "The text expresses amusement."
          },
          "id": "amusement",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "amusement",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The text does not express anger.",
            "true": "The text expresses anger."
          },
          "id": "anger",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "anger",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The text does not express annoyance.",
            "true": "The text expresses annoyance."
          },
          "id": "annoyance",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "annoyance",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The text does not express approval.",
            "true": "The text expresses approval."
          },
          "id": "approval",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "approval",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The text does not express caring.",
            "true": "The text expresses caring."
          },
          "id": "caring",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "caring",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The text does not express confusion.",
            "true": "The text expresses confusion."
          },
          "id": "confusion",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "confusion",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The text does not express curiosity.",
            "true": "The text expresses curiosity."
          },
          "id": "curiosity",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "curiosity",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The text does not express desire.",
            "true": "The text expresses desire."
          },
          "id": "desire",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "desire",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The text does not express disappointment.",
            "true": "The text expresses disappointment."
          },
          "id": "disappointment",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "disappointment",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The text does not express disapproval.",
            "true": "The text expresses disapproval."
          },
          "id": "disapproval",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "disapproval",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The text does not express disgust.",
            "true": "The text expresses disgust."
          },
          "id": "disgust",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "disgust",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The text does not express embarrassment.",
            "true": "The text expresses embarrassment."
          },
          "id": "embarrassment",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "embarrassment",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The text does not express excitement.",
            "true": "The text expresses excitement."
          },
          "id": "excitement",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "excitement",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The text does not express fear.",
            "true": "The text expresses fear."
          },
          "id": "fear",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "fear",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The text does not express gratitude.",
            "true": "The text expresses gratitude."
          },
          "id": "gratitude",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "gratitude",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The text does not express grief.",
            "true": "The text expresses grief."
          },
          "id": "grief",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "grief",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The text does not express joy.",
            "true": "The text expresses joy."
          },
          "id": "joy",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "joy",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The text does not express love.",
            "true": "The text expresses love."
          },
          "id": "love",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "love",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The text does not express nervousness.",
            "true": "The text expresses nervousness."
          },
          "id": "nervousness",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "nervousness",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The text expresses at least one identifiable emotion.",
            "true": "The text expresses no identifiable emotion."
          },
          "id": "neutral",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "neutral",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The text does not express optimism.",
            "true": "The text expresses optimism."
          },
          "id": "optimism",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "optimism",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The text does not express pride.",
            "true": "The text expresses pride."
          },
          "id": "pride",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "pride",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The text does not express realization.",
            "true": "The text expresses realization."
          },
          "id": "realization",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "realization",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The text does not express relief.",
            "true": "The text expresses relief."
          },
          "id": "relief",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "relief",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The text does not express remorse.",
            "true": "The text expresses remorse."
          },
          "id": "remorse",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "remorse",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The text does not express sadness.",
            "true": "The text expresses sadness."
          },
          "id": "sadness",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "sadness",
            "type": "noul"
          }
        },
        {
          "criteria_by_id": {
            "false": "The text does not express surprise.",
            "true": "The text expresses surprise."
          },
          "id": "surprise",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "The seven Civil Comments axes and all GoEmotions axes are independent judgments; retain each decision separately.",
          "selector": {
            "decision_id": "surprise",
            "type": "noul"
          }
        }
      ],
      "source_profile": "go_emotions"
    },
    "gretel_pii": {
      "disposition": "include",
      "disposition_reason": "See permissive dataset build note; short task-specific instructions, source-specific label checks and grouped splits. Author license overrides missing mirror metadata where documented.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/gretelai/gretel-pii-masking-en-v1/blob/e06eb1499ca8d54470f085021cd8e54f9efac7fd/README.md",
          "what_verified": "Upstream dataset/card identifier gretelai/gretel-pii-masking-en-v1 at pinned revision e06eb1499ca8d54470f085021cd8e54f9efac7fd; license metadata is ['apache-2.0']. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/gretelai/gretel-pii-masking-en-v1/raw/e06eb1499ca8d54470f085021cd8e54f9efac7fd/README.md",
          "what_verified": "Primary source card was fetched and read at e06eb1499ca8d54470f085021cd8e54f9efac7fd; card excerpt: # Gretel Synthetic Domain-Specific Documents Dataset (English) This dataset is a synthetically generated collection of documents enriched with Personally Identifiable Information (PII) and Protected Health Information (PHI) entities spanning multiple domains. Created using Gretel Navigator with mistral-nemo-2407 as the backend model, it is specifically designed for fine-tuning Gliner models. The dataset contains document passages featuring PII/PHI entities from a wide range of domain and documen"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/gretel_pii/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 3 source artifacts; 3 referenced local cache paths exist and integrity digests match. Plan source revision e06eb1499ca8d54470f085021cd8e54f9efac7fd."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/gretel_pii/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/gretel_pii",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 57887 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 9 actual rows from 3 pinned raw source files for source key gretel_pii; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "id": "gretel_pii__choice__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Preserve source Choice option IDs/descriptions and order; sampled option lists can be instance-specific.",
          "selector": {
            "decision_id": "mention_\\d+",
            "kind": "judgment",
            "type": "choice"
          }
        }
      ],
      "source_profile": "gretel_pii"
    },
    "gsm8k": {
      "disposition": "include",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching.",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "740312add88f781978c0658806c59bc2815b9866",
          "sha256": "a17e882503578c9e324560630e31017617714ee025c87cf4fea6fd916895f3c1",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/openai/gsm8k/740312add88f781978c0658806c59bc2815b9866/README.md?%2Fdatasets%2Fopenai%2Fgsm8k%2Fresolve%2F740312add88f781978c0658806c59bc2815b9866%2FREADME.md=&etag=%22ad25b2c451d0fcd2793678b2baa917c962acf281%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "gsm8k:main:test:0",
          "local_path": "output/typed_decisions/additional20-v1/cache/4b99860a39a9a702c6be204d0ba9ffac2d5bf2525c12d790f4263f46fa6b0844",
          "original_row_match_status": "matched",
          "sha256": "ee7b8da9e381df27b9e3f7758a159ab2bdaa4dbaa910546cbbc47e0cb44e4f59",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://huggingface.co/datasets/openai/gsm8k/resolve/a05f38c23a0e9ab0b71de8a2b4947e20f74f68f7/main/test/0000.parquet"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/additional_adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {},
          "id": "answer",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "Options and their descriptions vary by case; use the row’s own option-ID/description pairs. Do not assign a fixed meaning to option IDs across rows.",
          "option_semantics": "per_case_candidates",
          "preserve_case_options": true,
          "selector": {
            "decision_id": "answer",
            "type": "choice"
          }
        }
      ],
      "source_profile": "gsm8k"
    },
    "hans": {
      "disposition": "include",
      "disposition_reason": "See permissive dataset build note; short task-specific instructions, source-specific label checks and grouped splits. Author license overrides missing mirror metadata where documented.; For entailment-style Noul rows, false means the queried support/entailment relation is not established; do not relabel all false examples as contradictions.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/tasksource/hans/blob/8f107047d34fdfb57abfde38ec5196d63f7dee33/README.md",
          "what_verified": "Upstream dataset/card identifier tasksource/hans at pinned revision 8f107047d34fdfb57abfde38ec5196d63f7dee33; license metadata is ['mit']. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/tasksource/hans/raw/8f107047d34fdfb57abfde38ec5196d63f7dee33/README.md",
          "what_verified": "Primary source card was fetched and read at 8f107047d34fdfb57abfde38ec5196d63f7dee33; card excerpt: --- dataset_info: features: - name: sentence1 dtype: string - name: sentence2 dtype: string - name: labels"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/hans/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 3 source artifacts; 3 referenced local cache paths exist and integrity digests match. Plan source revision 8f107047d34fdfb57abfde38ec5196d63f7dee33."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/hans/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/hans",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 59806 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 9 actual rows from 3 pinned raw source files for source key hans; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "false": "The premise does not entail the hypothesis. This includes both contradiction and neutral cases.",
            "true": "The premise entails the hypothesis."
          },
          "id": "hans__noul__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Source-grounded false/true meanings; see source evidence.",
          "selector": {
            "decision_id": "entails",
            "kind": "judgment",
            "type": "noul"
          }
        }
      ],
      "source_profile": "hans"
    },
    "hatecheck": {
      "disposition": "include_with_caveat",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching. Caveats: Existing source spot-check note: \"機能テストのhate判定を確認。templateをgroupに用い、testのみ保持してtrainに混ぜない。\"",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "9d2ac89df04254e5c427bcc8d61b6d6c83a1f59b",
          "sha256": "1be194085db5f40457d8a63efb4f853a5d9a706fca9b9f77f0e95bc10601ce62",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/Paul/hatecheck/9d2ac89df04254e5c427bcc8d61b6d6c83a1f59b/README.md?%2Fdatasets%2FPaul%2Fhatecheck%2Fresolve%2F9d2ac89df04254e5c427bcc8d61b6d6c83a1f59b%2FREADME.md=&etag=%22c38f96474baefa34a38ee1e19fa148c7d5408ac7%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "hatecheck:default:test:0",
          "local_path": "output/typed_decisions/v1/cache/c73a1e07f8f67238614eddb15d35fc093652c80ec33eb1f325d9d7a9434c0bb6",
          "original_row_match_status": "matched",
          "sha256": "5fc0d2b011320a9a0ca84affb1b12fb8d38aff240162ecb0f44fac526fbfa68d",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://huggingface.co/datasets/Paul/hatecheck/resolve/5d64d7b973d8e11a3098ef5a7c25192cf1fefeee/default/test/0000.parquet"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/extra_adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "false": "The test case does not express hate toward a person or group based on identity.",
            "true": "The test case expresses hate toward a person or group based on identity."
          },
          "id": "hateful",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "",
          "selector": {
            "decision_id": "hateful",
            "type": "noul"
          }
        }
      ],
      "source_profile": "hatecheck"
    },
    "hh_rlhf": {
      "disposition": "include_with_caveat",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching. Caveats: Existing source spot-check note: \"helpful/harmlessを区別し、共通会話と最後の返答候補を分離。chosen側の名前を入力に出さない。選好の勝者でも不完全な返答があり、絶対品質の正例とは扱わない。\"",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "09be8c5bbc57cb3887f3a9732ad6aa7ec602a1fa",
          "sha256": "f75f40db0268656ba07736ec8e59a9720c1910ce554c85354cec74b1c8bda175",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/Anthropic/hh-rlhf/09be8c5bbc57cb3887f3a9732ad6aa7ec602a1fa/README.md?%2Fdatasets%2FAnthropic%2Fhh-rlhf%2Fresolve%2F09be8c5bbc57cb3887f3a9732ad6aa7ec602a1fa%2FREADME.md=&etag=%229fe5c91f53c12a62c1414e5d011085c747680d5f%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "hh_rlhf:harmless-base:test:0",
          "local_path": "output/typed_decisions/v1/cache/6a743154987bd6ee640d80a520f9e6052d7e266d760e84cc352aaa658fe466bf",
          "original_row_match_status": "matched",
          "sha256": "ebfaaed21162a4de120ae85366075b34425e2a6303fac59b99555002f7016f03",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://huggingface.co/datasets/Anthropic/hh-rlhf/resolve/09be8c5bbc57cb3887f3a9732ad6aa7ec602a1fa/harmless-base/test.jsonl.gz"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {},
          "id": "preferred_reply",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "Options and their descriptions vary by case; use the row’s own option-ID/description pairs. Do not assign a fixed meaning to option IDs across rows.",
          "option_semantics": "per_case_candidates",
          "preserve_case_options": true,
          "selector": {
            "decision_id": "preferred_reply",
            "type": "choice"
          }
        }
      ],
      "source_profile": "hh_rlhf"
    },
    "hwu64": {
      "disposition": "include_with_caveat",
      "disposition_reason": "FastFit mirror retains train/validation/test. CC-BY-SA 3.0 per DialoGLUE; do not infer a new license from unlicensed mirror.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/FastFit/hwu_64/blob/9fb5bff1e37c4c5a6d46c1ff6f286b6d3f362222/README.md",
          "what_verified": "Upstream dataset/card identifier FastFit/hwu_64 at pinned revision 9fb5bff1e37c4c5a6d46c1ff6f286b6d3f362222; license metadata is ['cc-by-sa-3.0']. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/FastFit/hwu_64/raw/9fb5bff1e37c4c5a6d46c1ff6f286b6d3f362222/README.md",
          "what_verified": "Primary source card was fetched and read at 9fb5bff1e37c4c5a6d46c1ff6f286b6d3f362222; card excerpt: --- dataset_info: features: - name: label dtype: string - name: text dtype: string splits:"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/hwu64/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 3 source artifacts; 3 referenced local cache paths exist and integrity digests match. Plan source revision 9fb5bff1e37c4c5a6d46c1ff6f286b6d3f362222."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/hwu64/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/hwu64",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 9333 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 9 actual rows from 3 pinned raw source files for source key hwu64; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "id": "hwu64__choice__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Preserve source Choice option IDs/descriptions and order; sampled option lists can be instance-specific.",
          "selector": {
            "decision_id": "label",
            "kind": "judgment",
            "type": "choice"
          }
        }
      ],
      "source_profile": "hwu64"
    },
    "impli": {
      "disposition": "include",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching.",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "d43c272c895c773b311018d51ab65f592c6699bb",
          "sha256": "4399ed6a2d9e24071fa5378a52f33b7270dddf20805315b4d425ee3cfa7e52bf",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://raw.githubusercontent.com/UKPLab/acl2022-impli/d43c272c895c773b311018d51ab65f592c6699bb/README.md"
        },
        {
          "adapter_replay": "matched",
          "case_id": "impli:idioms/manual_antonyms_ne:test:0",
          "local_path": "output/typed_decisions/additional20-v1/cache/a3d094780efcf1ebbf80da36aad015944f9fd557af27fdcefd93097aa3d6017a",
          "original_row_match_status": "matched",
          "sha256": "f7803d1cdcc6f7af00a9c9a7e2bcc7fd96206f7f888ab8b98950a78247de3ce2",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://raw.githubusercontent.com/UKPLab/acl2022-impli/d43c272c895c773b311018d51ab65f592c6699bb/idioms/manual_antonyms_ne.tsv"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/additional_adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "false": "The premise does not imply the hypothesis in its intended meaning.",
            "true": "The premise implies the hypothesis in its intended meaning."
          },
          "id": "entails",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "",
          "selector": {
            "decision_id": "entails",
            "type": "noul"
          }
        }
      ],
      "source_profile": "impli"
    },
    "lexcomp": {
      "disposition": "include",
      "disposition_reason": "See permissive dataset build note; short task-specific instructions, source-specific label checks and grouped splits. Author license overrides missing mirror metadata where documented.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/tasksource/lexcomp-nc-relation/blob/4f7fa5d487d71ac1543bc7849993b5dfc2e361ac/README.md",
          "what_verified": "Upstream dataset/card identifier tasksource/lexcomp-nc-relation at pinned revision 4f7fa5d487d71ac1543bc7849993b5dfc2e361ac; license metadata is ['apache-2.0']. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/tasksource/lexcomp-nc-relation/raw/4f7fa5d487d71ac1543bc7849993b5dfc2e361ac/README.md",
          "what_verified": "Primary source card was fetched and read at 4f7fa5d487d71ac1543bc7849993b5dfc2e361ac; card excerpt: --- license: apache-2.0 language: - en --- https://github.com/vered1986/lexcomp/tree/master ``` @article{shwartz-dagan-2019-still,"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/lexcomp/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 3 source artifacts; 3 referenced local cache paths exist and integrity digests match. Plan source revision 4f7fa5d487d71ac1543bc7849993b5dfc2e361ac."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/lexcomp/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/lexcomp",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 1564 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 9 actual rows from 3 pinned raw source files for source key lexcomp; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "false": "The paraphrase does not describe the compound’s meaning in the given context.",
            "true": "The paraphrase describes the compound’s meaning in the given context."
          },
          "id": "lexcomp__noul__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Source-grounded false/true meanings; see source evidence.",
          "selector": {
            "decision_id": "meaning",
            "kind": "judgment",
            "type": "noul"
          }
        }
      ],
      "source_profile": "lexcomp"
    },
    "logical_entailment": {
      "disposition": "include",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching.",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "5275a96cea765dc3bbfd41e6034485791704ed69",
          "sha256": "2339009bf80882a67e61cf2a9cd8b4716cfeff5b7faa5c285e3754548c729f98",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/tasksource/logical-entailment/5275a96cea765dc3bbfd41e6034485791704ed69/README.md?%2Fdatasets%2Ftasksource%2Flogical-entailment%2Fresolve%2F5275a96cea765dc3bbfd41e6034485791704ed69%2FREADME.md=&etag=%22e47865dbf2b344cfd415b680cfbdc696e9433739%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "logical_entailment:default:test:0",
          "local_path": "output/typed_decisions/additional20-v1/cache/66f7f765922d94436191f3ee96ef3dc7e6adaa7cbef2e3f7163b7d6f07ac5a5c",
          "original_row_match_status": "matched",
          "sha256": "221df6c5f506b89f9d58cd62819382bd7b37514b2b5f15116932f98c557c244e",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://huggingface.co/datasets/tasksource/logical-entailment/resolve/90d8d477b1b7f6bc4ba05ea26b84574763fa2637/default/test/0000.parquet"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/additional_adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "false": "The logical premises do not entail the conclusion; this does not by itself establish that the conclusion is false.",
            "true": "The logical premises entail the conclusion."
          },
          "id": "entails",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "",
          "selector": {
            "decision_id": "entails",
            "type": "noul"
          }
        }
      ],
      "source_profile": "logical_entailment"
    },
    "lonli": {
      "disposition": "include_with_caveat",
      "disposition_reason": "See permissive dataset build note; short task-specific instructions, source-specific label checks and grouped splits. Author license overrides missing mirror metadata where documented.; No supervised training split exists in this release subset; treat available rows as evaluation-only/source-limited.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/tasksource/lonli/blob/f17e55b8f60da2f32720c1a16e4b6149145c662f/README.md",
          "what_verified": "Upstream dataset/card identifier tasksource/lonli at pinned revision f17e55b8f60da2f32720c1a16e4b6149145c662f; license metadata is ['mit']. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/tasksource/lonli/raw/f17e55b8f60da2f32720c1a16e4b6149145c662f/README.md",
          "what_verified": "Primary source card was fetched and read at f17e55b8f60da2f32720c1a16e4b6149145c662f; card excerpt: --- license: mit task_ids: - natural-language-inference task_categories: - text-classification language: - en"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/lonli/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 1 source artifacts; 1 referenced local cache paths exist and integrity digests match. Plan source revision f17e55b8f60da2f32720c1a16e4b6149145c662f."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/lonli/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/lonli",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 348886 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 3 actual rows from 1 pinned raw source files for source key lonli; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "id": "lonli__choice__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Preserve source Choice option IDs/descriptions and order; sampled option lists can be instance-specific.",
          "selector": {
            "decision_id": "relation",
            "kind": "judgment",
            "type": "choice"
          }
        }
      ],
      "source_profile": "lonli"
    },
    "massive": {
      "disposition": "include_with_caveat",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching. Caveats: Existing source spot-check note: \"playとmusicの短いラベルだけでは紛らわしいためdomain説明を具体化。混合全体でCLINC評価発話と一致するtrain42件を除外し再確認。 追加後のtrainサンプルに delete this email → email sendemail があり、sourceの分類体系と直観のずれにも注意。\"",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "ff6bd8e4b27c3543e4f8fe2108f32bb95a6f8740",
          "sha256": "9a74b2298db2cb509eb39725612ef5d5b9eb99d59ef7c960ab01f58526051e3d",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/AmazonScience/massive/ff6bd8e4b27c3543e4f8fe2108f32bb95a6f8740/README.md?%2Fdatasets%2FAmazonScience%2Fmassive%2Fresolve%2Fff6bd8e4b27c3543e4f8fe2108f32bb95a6f8740%2FREADME.md=&etag=%22acee1cd9ccd5cdc6af177d9c9db747265e17a5c9%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "massive:en-US:test:0",
          "local_path": "output/typed_decisions/v1/cache/964b44c4ddb47581fdeb5c840578005121c0f9d14e0d90512c72298e76d25198",
          "original_row_match_status": "matched",
          "sha256": "4cba5faa11c71437928e17cb1b9b3d8b8e727e7ea363a3a9a8045e19c0491577",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://amazon-massive-nlu-dataset.s3.amazonaws.com/amazon-massive-dataset-1.1.tar.gz"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "alarm_query": "alarm query",
            "alarm_remove": "alarm remove",
            "alarm_set": "alarm set",
            "audio_volume_down": "audio volume down",
            "audio_volume_mute": "audio volume mute",
            "audio_volume_other": "audio volume other",
            "audio_volume_up": "audio volume up",
            "calendar_query": "calendar query",
            "calendar_remove": "calendar remove",
            "calendar_set": "calendar set",
            "cooking_query": "cooking query",
            "cooking_recipe": "cooking recipe",
            "datetime_convert": "datetime convert",
            "datetime_query": "datetime query",
            "email_addcontact": "email addcontact",
            "email_query": "email query",
            "email_querycontact": "email querycontact",
            "email_sendemail": "email sendemail",
            "general_greet": "general greet",
            "general_joke": "general joke",
            "general_quirky": "general quirky",
            "iot_cleaning": "iot cleaning",
            "iot_coffee": "iot coffee",
            "iot_hue_lightchange": "iot hue lightchange",
            "iot_hue_lightdim": "iot hue lightdim",
            "iot_hue_lightoff": "iot hue lightoff",
            "iot_hue_lighton": "iot hue lighton",
            "iot_hue_lightup": "iot hue lightup",
            "iot_wemo_off": "iot wemo off",
            "iot_wemo_on": "iot wemo on",
            "lists_createoradd": "lists createoradd",
            "lists_query": "lists query",
            "lists_remove": "lists remove",
            "music_dislikeness": "music dislikeness",
            "music_likeness": "music likeness",
            "music_query": "music query",
            "music_settings": "music settings",
            "news_query": "news query",
            "play_audiobook": "play audiobook",
            "play_game": "play game",
            "play_music": "play music",
            "play_podcasts": "play podcasts",
            "play_radio": "play radio",
            "qa_currency": "qa currency",
            "qa_definition": "qa definition",
            "qa_factoid": "qa factoid",
            "qa_maths": "qa maths",
            "qa_stock": "qa stock",
            "recommendation_events": "recommendation events",
            "recommendation_locations": "recommendation locations",
            "recommendation_movies": "recommendation movies",
            "social_post": "social post",
            "social_query": "social query",
            "takeaway_order": "takeaway order",
            "takeaway_query": "takeaway query",
            "transport_query": "transport query",
            "transport_taxi": "transport taxi",
            "transport_ticket": "transport ticket",
            "transport_traffic": "transport traffic",
            "weather_query": "weather query"
          },
          "id": "intent",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "Descriptions are copied from the pinned source conversion’s case criteria. Keep IDs aligned; only expand them from source definitions, never from the observed target.",
          "option_semantics": "fixed_source_taxonomy",
          "selector": {
            "decision_id": "intent",
            "type": "choice"
          }
        },
        {
          "criteria_by_id": {
            "alarm": "alarm: alarm query, alarm remove, alarm set",
            "audio": "audio: audio volume down, audio volume mute, audio volume other, audio volume up",
            "calendar": "calendar: calendar query, calendar remove, calendar set",
            "cooking": "cooking: cooking query, cooking recipe",
            "datetime": "datetime: datetime convert, datetime query",
            "email": "email: email addcontact, email query, email querycontact, email sendemail",
            "general": "general: general greet, general joke, general quirky",
            "iot": "iot: iot cleaning, iot coffee, iot hue lightchange, iot hue lightdim, iot hue lightoff, iot hue lighton, iot hue lightup, iot wemo off, iot wemo on",
            "lists": "lists: lists createoradd, lists query, lists remove",
            "music": "Music preferences, settings, likes, or music information.",
            "news": "news: news query",
            "play": "Playback requests for music, radio, audiobooks, or games.",
            "qa": "qa: qa currency, qa definition, qa factoid, qa maths, qa stock",
            "recommendation": "recommendation: recommendation events, recommendation locations, recommendation movies",
            "social": "social: social post, social query",
            "takeaway": "takeaway: takeaway order, takeaway query",
            "transport": "transport: transport query, transport taxi, transport ticket, transport traffic",
            "weather": "weather: weather query"
          },
          "id": "scenario",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "Descriptions are copied from the pinned source conversion’s case criteria. Keep IDs aligned; only expand them from source definitions, never from the observed target.",
          "option_semantics": "fixed_source_taxonomy",
          "selector": {
            "decision_id": "scenario",
            "type": "choice"
          }
        }
      ],
      "source_profile": "massive"
    },
    "miqa": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Original google-research/language data under Apache-2.0; tasksource distribution declares CC-BY-4.0. Retain both notices. Forward inference pairs only; evaluation-only.; No supervised training split exists in this release subset; treat available rows as evaluation-only/source-limited.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/tasksource/miqa/blob/b0b4a19860eea1740f04a7836612ec239e08e838/README.md",
          "what_verified": "Upstream dataset/card identifier tasksource/miqa at pinned revision b0b4a19860eea1740f04a7836612ec239e08e838; license metadata is ['apache-2.0', 'cc-by-4.0']. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/tasksource/miqa/raw/b0b4a19860eea1740f04a7836612ec239e08e838/README.md",
          "what_verified": "Primary source card was fetched and read at b0b4a19860eea1740f04a7836612ec239e08e838; card excerpt: --- license: cc-by-4.0 ---"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/miqa/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 1 source artifacts; 1 referenced local cache paths exist and integrity digests match. Plan source revision b0b4a19860eea1740f04a7836612ec239e08e838."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/miqa/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/miqa",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 300 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 3 actual rows from 1 pinned raw source files for source key miqa; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "id": "miqa__choice__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Preserve source Choice option IDs/descriptions and order; sampled option lists can be instance-specific.",
          "selector": {
            "decision_id": "inference",
            "kind": "judgment",
            "type": "choice"
          }
        }
      ],
      "source_profile": "miqa"
    },
    "mtop": {
      "disposition": "include_with_caveat",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching. Caveats: Existing source spot-check note: \"英語en_XXだけを取り、intentとdomainを別Choiceとして確認。gold logical formは入力へ出さない。\"",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "9fa95557ba2809ad05943880eee5231b4f2463fb",
          "sha256": "414e2ce25ba38c5cd0884dcd74d9893812a34f3b1491f6d63d7d89e49fc8a690",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/tasksource/mtop/9fa95557ba2809ad05943880eee5231b4f2463fb/README.md?%2Fdatasets%2Ftasksource%2Fmtop%2Fresolve%2F9fa95557ba2809ad05943880eee5231b4f2463fb%2FREADME.md=&etag=%228166eea596c21c135aa28985a34719b280fde365%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "mtop:default:test:3549",
          "local_path": "output/typed_decisions/v1/cache/91ec8f30058610fa3b9d741da53956e7551042d9889fb17f6731873287e10439",
          "original_row_match_status": "matched",
          "sha256": "d9a2a2784fb199f45a78067b6567daada8c4e641c7bdb4afbcb1dc5579ef5ccd",
          "source": "pinned_raw_example",
          "source_row_index": 3549,
          "state_and_decisions_matched": true,
          "url": "https://huggingface.co/datasets/tasksource/mtop/resolve/672acca15e7a6863506569bc610daed0a4416e1e/default/test/0000.parquet"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/extra_adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "alarm": "alarm",
            "calling": "calling",
            "event": "event",
            "messaging": "messaging",
            "music": "music",
            "news": "news",
            "people": "people",
            "recipes": "recipes",
            "reminder": "reminder",
            "timer": "timer",
            "weather": "weather"
          },
          "id": "domain",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "Descriptions are copied from the pinned source conversion’s case criteria. Keep IDs aligned; only expand them from source definitions, never from the observed target.",
          "option_semantics": "fixed_source_taxonomy",
          "selector": {
            "decision_id": "domain",
            "type": "choice"
          }
        },
        {
          "criteria_by_id": {
            "IN:ADD_TIME_TIMER": "add time timer",
            "IN:ADD_TO_PLAYLIST_MUSIC": "add to playlist music",
            "IN:ANSWER_CALL": "answer call",
            "IN:CANCEL_CALL": "cancel call",
            "IN:CANCEL_MESSAGE": "cancel message",
            "IN:CREATE_ALARM": "create alarm",
            "IN:CREATE_CALL": "create call",
            "IN:CREATE_PLAYLIST_MUSIC": "create playlist music",
            "IN:CREATE_REMINDER": "create reminder",
            "IN:CREATE_TIMER": "create timer",
            "IN:DELETE_ALARM": "delete alarm",
            "IN:DELETE_PLAYLIST_MUSIC": "delete playlist music",
            "IN:DELETE_REMINDER": "delete reminder",
            "IN:DELETE_TIMER": "delete timer",
            "IN:DISLIKE_MUSIC": "dislike music",
            "IN:DISPREFER": "disprefer",
            "IN:END_CALL": "end call",
            "IN:FAST_FORWARD_MUSIC": "fast forward music",
            "IN:FOLLOW_MUSIC": "follow music",
            "IN:GET_AGE": "get age",
            "IN:GET_AIRQUALITY": "get airquality",
            "IN:GET_ALARM": "get alarm",
            "IN:GET_ATTENDEE_EVENT": "get attendee event",
            "IN:GET_AVAILABILITY": "get availability",
            "IN:GET_CALL": "get call",
            "IN:GET_CALL_CONTACT": "get call contact",
            "IN:GET_CALL_TIME": "get call time",
            "IN:GET_CATEGORY_EVENT": "get category event",
            "IN:GET_CONTACT": "get contact",
            "IN:GET_CONTACT_METHOD": "get contact method",
            "IN:GET_DATE_TIME_EVENT": "get date time event",
            "IN:GET_DETAILS_NEWS": "get details news",
            "IN:GET_EDUCATION_DEGREE": "get education degree",
            "IN:GET_EDUCATION_TIME": "get education time",
            "IN:GET_EMPLOYER": "get employer",
            "IN:GET_EMPLOYMENT_TIME": "get employment time",
            "IN:GET_EVENT": "get event",
            "IN:GET_GENDER": "get gender",
            "IN:GET_GROUP": "get group",
            "IN:GET_INFO_CONTACT": "get info contact",
            "IN:GET_INFO_RECIPES": "get info recipes",
            "IN:GET_JOB": "get job",
            "IN:GET_LANGUAGE": "get language",
            "IN:GET_LIFE_EVENT": "get life event",
            "IN:GET_LIFE_EVENT_TIME": "get life event time",
            "IN:GET_LOCATION": "get location",
            "IN:GET_LYRICS_MUSIC": "get lyrics music",
            "IN:GET_MAJOR": "get major",
            "IN:GET_MESSAGE": "get message",
            "IN:GET_MESSAGE_CONTACT": "get message contact",
            "IN:GET_MUTUAL_FRIENDS": "get mutual friends",
            "IN:GET_RECIPES": "get recipes",
            "IN:GET_REMINDER": "get reminder",
            "IN:GET_REMINDER_AMOUNT": "get reminder amount",
            "IN:GET_REMINDER_DATE_TIME": "get reminder date time",
            "IN:GET_REMINDER_LOCATION": "get reminder location",
            "IN:GET_STORIES_NEWS": "get stories news",
            "IN:GET_SUNRISE": "get sunrise",
            "IN:GET_SUNSET": "get sunset",
            "IN:GET_TIMER": "get timer",
            "IN:GET_TRACK_INFO_MUSIC": "get track info music",
            "IN:GET_UNDERGRAD": "get undergrad",
            "IN:GET_WEATHER": "get weather",
            "IN:HELP_REMINDER": "help reminder",
            "IN:HOLD_CALL": "hold call",
            "IN:IGNORE_CALL": "ignore call",
            "IN:IS_TRUE_RECIPES": "is true recipes",
            "IN:LIKE_MUSIC": "like music",
            "IN:LOOP_MUSIC": "loop music",
            "IN:MERGE_CALL": "merge call",
            "IN:PAUSE_MUSIC": "pause music",
            "IN:PAUSE_TIMER": "pause timer",
            "IN:PLAY_MEDIA": "play media",
            "IN:PLAY_MUSIC": "play music",
            "IN:PREFER": "prefer",
            "IN:PREVIOUS_TRACK_MUSIC": "previous track music",
            "IN:QUESTION_MUSIC": "question music",
            "IN:QUESTION_NEWS": "question news",
            "IN:REMOVE_FROM_PLAYLIST_MUSIC": "remove from playlist music",
            "IN:REPEAT_ALL_MUSIC": "repeat all music",
            "IN:REPEAT_ALL_OFF_MUSIC": "repeat all off music",
            "IN:REPLAY_MUSIC": "replay music",
            "IN:RESTART_TIMER": "restart timer",
            "IN:RESUME_CALL": "resume call",
            "IN:RESUME_MUSIC": "resume music",
            "IN:RESUME_TIMER": "resume timer",
            "IN:REWIND_MUSIC": "rewind music",
            "IN:SEND_MESSAGE": "send message",
            "IN:SET_AVAILABLE": "set available",
            "IN:SET_DEFAULT_PROVIDER_CALLING": "set default provider calling",
            "IN:SET_DEFAULT_PROVIDER_MUSIC": "set default provider music",
            "IN:SET_RSVP_INTERESTED": "set rsvp interested",
            "IN:SET_RSVP_NO": "set rsvp no",
            "IN:SET_RSVP_YES": "set rsvp yes",
            "IN:SET_UNAVAILABLE": "set unavailable",
            "IN:SHARE_EVENT": "share event",
            "IN:SILENCE_ALARM": "silence alarm",
            "IN:SKIP_TRACK_MUSIC": "skip track music",
            "IN:SNOOZE_ALARM": "snooze alarm",
            "IN:START_SHUFFLE_MUSIC": "start shuffle music",
            "IN:STOP_MUSIC": "stop music",
            "IN:STOP_SHUFFLE_MUSIC": "stop shuffle music",
            "IN:SUBTRACT_TIME_TIMER": "subtract time timer",
            "IN:SWITCH_CALL": "switch call",
            "IN:UNLOOP_MUSIC": "unloop music",
            "IN:UPDATE_ALARM": "update alarm",
            "IN:UPDATE_CALL": "update call",
            "IN:UPDATE_METHOD_CALL": "update method call",
            "IN:UPDATE_REMINDER": "update reminder",
            "IN:UPDATE_REMINDER_DATE_TIME": "update reminder date time",
            "IN:UPDATE_REMINDER_LOCATION": "update reminder location",
            "IN:UPDATE_REMINDER_TODO": "update reminder todo",
            "IN:UPDATE_TIMER": "update timer"
          },
          "id": "intent",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "Descriptions are copied from the pinned source conversion’s case criteria. Keep IDs aligned; only expand them from source definitions, never from the observed target.",
          "option_semantics": "fixed_source_taxonomy",
          "selector": {
            "decision_id": "intent",
            "type": "choice"
          }
        }
      ],
      "source_profile": "mtop"
    },
    "nlsat": {
      "disposition": "include",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching.",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "1c7e1a3afbb62462ee979d8d77521ebbfcef0255",
          "sha256": "98b45ea81164d1e1a1dd82255207053b15cd6c69d922a1c5cf3387ce604d4b74",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/tasksource/nlsat/1c7e1a3afbb62462ee979d8d77521ebbfcef0255/README.md?%2Fdatasets%2Ftasksource%2Fnlsat%2Fresolve%2F1c7e1a3afbb62462ee979d8d77521ebbfcef0255%2FREADME.md=&etag=%22154df8298fab5ecf322016157858e08cd1bccbe1%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "nlsat:default:test:0",
          "local_path": "output/typed_decisions/permissive-v1/cache/5760c753a25bc83395405419d88045f61c6eb95365d948f051519ccdf39298d0",
          "original_row_match_status": "matched",
          "sha256": "58c49cd84af82bfe23368b6195c31d0cf96e967871dfda180feed5bfd7f42a4a",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://huggingface.co/datasets/tasksource/nlsat/resolve/1c7e1a3afbb62462ee979d8d77521ebbfcef0255/test.jsonl"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/permissive_adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "false": "The rules and facts in the state cannot all be true at the same time.",
            "true": "All rules and facts in the state can be true at the same time."
          },
          "id": "satisfiable",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "",
          "selector": {
            "decision_id": "satisfiable",
            "type": "noul"
          }
        }
      ],
      "source_profile": "nlsat"
    },
    "openbookqa": {
      "disposition": "include",
      "disposition_reason": "See permissive dataset build note; short task-specific instructions, source-specific label checks and grouped splits. Author license overrides missing mirror metadata where documented.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/allenai/openbookqa/blob/388097ea7776314e93a529163e0fea805b8a6454/README.md",
          "what_verified": "Upstream dataset/card identifier allenai/openbookqa at pinned revision 388097ea7776314e93a529163e0fea805b8a6454; license metadata is ['apache-2.0']. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/allenai/openbookqa/raw/388097ea7776314e93a529163e0fea805b8a6454/README.md",
          "what_verified": "Primary source card was fetched and read at 388097ea7776314e93a529163e0fea805b8a6454; card excerpt: # Dataset Card for OpenBookQA ## Table of Contents - [Dataset Description](#dataset-description) - [Dataset Summary](#dataset-summary)"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/openbookqa/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 3 source artifacts; 3 referenced local cache paths exist and integrity digests match. Plan source revision 388097ea7776314e93a529163e0fea805b8a6454."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/openbookqa/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/openbookqa",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 5870 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 9 actual rows from 3 pinned raw source files for source key openbookqa; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "id": "openbookqa__choice__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Preserve source Choice option IDs/descriptions and order; sampled option lists can be instance-specific.",
          "selector": {
            "decision_id": "answer",
            "kind": "judgment",
            "type": "choice"
          }
        }
      ],
      "source_profile": "openbookqa"
    },
    "patent_similarity": {
      "disposition": "include",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching.",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "ff1868796e44ac9ec02e870b447fde6c5bf2d6e4",
          "sha256": "e2f7c8a58620aad6d290236e084bffa25aa0d5e514216686c0a5ca4aed09e22f",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/tasksource/patent-phrase-similarity/ff1868796e44ac9ec02e870b447fde6c5bf2d6e4/README.md?%2Fdatasets%2Ftasksource%2Fpatent-phrase-similarity%2Fresolve%2Fff1868796e44ac9ec02e870b447fde6c5bf2d6e4%2FREADME.md=&etag=%2295eb9984a4b30f08c42a8f4f684eb5d95d4d5d5a%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "patent_similarity:default:test:0",
          "local_path": "output/typed_decisions/permissive-v1/cache/95e9d4806c9487e0190a0bbb195393815c4902556e2820887548414f2817c969",
          "original_row_match_status": "matched",
          "sha256": "583014566ba7c9f967cef7cacb93ed8e5de8cf04c1fcad9a95e21e0839d0fdfc",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://huggingface.co/datasets/tasksource/patent-phrase-similarity/resolve/a4839e4f787f59675baf71c70dba8d078e1468c2/default/test/0000.parquet"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/permissive_adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {},
          "id": "similarity",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "Keep every ordered source level and the per-row source rubric as supplied. Do not replace or renumber it.",
          "preserve_source_score_rubric": true,
          "selector": {
            "decision_id": "similarity",
            "type": "score"
          }
        }
      ],
      "source_profile": "patent_similarity"
    },
    "paws": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Use PAWS-Wiki labeled_final only. Card licensing text permits any purpose; do not infer QQP rights. Unlabeled data is noisy.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/google-research-datasets/paws/blob/161ece9501cf0a11f3e48bd356eaa82de46d6a09/README.md",
          "what_verified": "Upstream dataset/card identifier google-research-datasets/paws at pinned revision 161ece9501cf0a11f3e48bd356eaa82de46d6a09; license metadata is ['other']. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/google-research-datasets/paws/raw/161ece9501cf0a11f3e48bd356eaa82de46d6a09/README.md",
          "what_verified": "Primary source card was fetched and read at 161ece9501cf0a11f3e48bd356eaa82de46d6a09; card excerpt: # Dataset Card for PAWS: Paraphrase Adversaries from Word Scrambling ## Table of Contents - [Dataset Description](#dataset-description) - [Dataset Summary](#dataset-summary)"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/paws/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 3 source artifacts; 3 referenced local cache paths exist and integrity digests match. Plan source revision 161ece9501cf0a11f3e48bd356eaa82de46d6a09."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/paws/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/paws",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 65305 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 9 actual rows from 3 pinned raw source files for source key paws; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "false": "The two sentences do not express the same meaning.",
            "true": "The two sentences express the same meaning."
          },
          "id": "paws__noul__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Source-grounded false/true meanings; see source evidence.",
          "selector": {
            "decision_id": "paraphrase",
            "kind": "judgment",
            "type": "noul"
          }
        }
      ],
      "source_profile": "paws"
    },
    "plane": {
      "disposition": "include",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching.",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "5ca649ccef611c6d2b453d4b34d25477a63bd9e4",
          "sha256": "3a33cfa8a94d2bb054fa706d28e4b765cd69c444af10836651297ba9d50a805a",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/tasksource/PLANE-ood/5ca649ccef611c6d2b453d4b34d25477a63bd9e4/README.md?%2Fdatasets%2Ftasksource%2FPLANE-ood%2Fresolve%2F5ca649ccef611c6d2b453d4b34d25477a63bd9e4%2FREADME.md=&etag=%223491029d5a34166cb1de488d31ceda2391fd388e%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "plane:default:test:0",
          "local_path": "output/typed_decisions/permissive-v1/cache/8c070a2255333813f91d701150ca439736c50da0d7a24e86eef294d7028d17b3",
          "original_row_match_status": "matched",
          "sha256": "02cbdb0cbfc0763cc65a3c2ec212ada98754d418ea7382e2e39b0f6fc6dfd27c",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://huggingface.co/datasets/tasksource/PLANE-ood/resolve/73af27c12a6ff5903084aa5fbcf4684c01741c55/default/test/0000.parquet"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/permissive_adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "false": "The category inference does not follow from the premise when the adjective’s effect on the noun is taken into account.",
            "true": "The category inference follows from the premise, accounting for the adjective’s effect on the noun."
          },
          "id": "entails",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "",
          "selector": {
            "decision_id": "entails",
            "type": "noul"
          }
        }
      ],
      "source_profile": "plane"
    },
    "poem_sentiment": {
      "disposition": "include",
      "disposition_reason": "See permissive dataset build note; short task-specific instructions, source-specific label checks and grouped splits. Author license overrides missing mirror metadata where documented.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/google-research-datasets/poem_sentiment/blob/685b95a2787a869b7bae6c4480810f57fe23b48e/README.md",
          "what_verified": "Upstream dataset/card identifier google-research-datasets/poem_sentiment at pinned revision 685b95a2787a869b7bae6c4480810f57fe23b48e; license metadata is ['cc-by-4.0']. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/google-research-datasets/poem_sentiment/raw/685b95a2787a869b7bae6c4480810f57fe23b48e/README.md",
          "what_verified": "Primary source card was fetched and read at 685b95a2787a869b7bae6c4480810f57fe23b48e; card excerpt: # Dataset Card for Gutenberg Poem Dataset ## Table of Contents - [Dataset Description](#dataset-description) - [Dataset Summary](#dataset-summary)"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/poem_sentiment/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 3 source artifacts; 3 referenced local cache paths exist and integrity digests match. Plan source revision 685b95a2787a869b7bae6c4480810f57fe23b48e."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/poem_sentiment/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/poem_sentiment",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 1101 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 9 actual rows from 3 pinned raw source files for source key poem_sentiment; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "id": "poem_sentiment__choice__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Preserve source Choice option IDs/descriptions and order; sampled option lists can be instance-specific.",
          "selector": {
            "decision_id": "sentiment",
            "kind": "judgment",
            "type": "choice"
          }
        }
      ],
      "source_profile": "poem_sentiment"
    },
    "qasper": {
      "disposition": "include",
      "disposition_reason": "See permissive dataset build note; short task-specific instructions, source-specific label checks and grouped splits. Author license overrides missing mirror metadata where documented.; Only unanimous explicit yes/no source annotations are retained; this is a source-annotated yes/no answer judgment, not a generic answerability target or a generated answer quality score.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/allenai/qasper/blob/fdc9d8214fbab5dd782958601db4d678e6934a54/README.md",
          "what_verified": "Upstream dataset/card identifier allenai/qasper at pinned revision fdc9d8214fbab5dd782958601db4d678e6934a54; license metadata is ['cc-by-4.0']. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/allenai/qasper/raw/fdc9d8214fbab5dd782958601db4d678e6934a54/README.md",
          "what_verified": "Primary source card was fetched and read at fdc9d8214fbab5dd782958601db4d678e6934a54; card excerpt: # Dataset Card for Qasper ## Table of Contents - [Table of Contents](#table-of-contents) - [Dataset Description](#dataset-description)"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/qasper/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 3 source artifacts; 3 referenced local cache paths exist and integrity digests match. Plan source revision fdc9d8214fbab5dd782958601db4d678e6934a54."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/qasper/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/qasper",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 607 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 9 actual rows from 3 pinned raw source files for source key qasper; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "false": "The paper’s answer to the question is no.",
            "true": "The paper’s answer to the question is yes."
          },
          "id": "qasper__noul__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Source-grounded false/true meanings; see source evidence.",
          "selector": {
            "decision_id": "answer",
            "kind": "judgment",
            "type": "noul"
          }
        }
      ],
      "source_profile": "qasper"
    },
    "quartz": {
      "disposition": "include_with_caveat",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching. Caveats: Existing source spot-check note: \"backgroundを落とさず、大小・増減の候補と正解を確認。元質問の表現が粗い例はそのまま保持。\"",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "28c1dbb56caf81799296cb17892fa73402e23464",
          "sha256": "0a7279210dc1c2310f5f0972d09058423c81a051a4fab8fbf8255e3be66ebb67",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/allenai/quartz/28c1dbb56caf81799296cb17892fa73402e23464/README.md?%2Fdatasets%2Fallenai%2Fquartz%2Fresolve%2F28c1dbb56caf81799296cb17892fa73402e23464%2FREADME.md=&etag=%22145e7d95758e28f096dc7faabe0d7781162811dc%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "quartz:default:test:0",
          "local_path": "output/typed_decisions/v1/cache/6f067b78114e38c1b7c8a6620d3d44fcdea21811284fab4c30a78223ede27c8a",
          "original_row_match_status": "matched",
          "sha256": "78c4c77cfc1ad0e65dbdfeec4ed788117e73484ad44f6a2aebf43e08c7fba2e8",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://huggingface.co/datasets/allenai/quartz/resolve/b041d210150e802ed192b3b84037c3a9c4542fc9/default/test/0000.parquet"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {},
          "id": "answer",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "Options and their descriptions vary by case; use the row’s own option-ID/description pairs. Do not assign a fixed meaning to option IDs across rows.",
          "option_semantics": "per_case_candidates",
          "preserve_case_options": true,
          "selector": {
            "decision_id": "answer",
            "type": "choice"
          }
        }
      ],
      "source_profile": "quartz"
    },
    "robust_lr": {
      "disposition": "include_with_caveat",
      "disposition_reason": "See insights/2026-09-24-additional20-dataset-build.md for source license, construction and typed-label review.; No supervised training split exists in this release subset; treat available rows as evaluation-only/source-limited.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/tasksource/robustLR/blob/369ce16fbe4467f46916efe3730d04fb607c0391/README.md",
          "what_verified": "Upstream dataset/card identifier tasksource/robustLR at pinned revision 369ce16fbe4467f46916efe3730d04fb607c0391; license metadata is ['mit']. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/tasksource/robustLR/raw/369ce16fbe4467f46916efe3730d04fb607c0391/README.md",
          "what_verified": "Primary source card was fetched and read at 369ce16fbe4467f46916efe3730d04fb607c0391; card excerpt: # Dataset Card for \"robustLR\" https://github.com/INK-USC/RobustLR ```"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/robust_lr/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 3 source artifacts; 3 referenced local cache paths exist and integrity digests match. Plan source revision 369ce16fbe4467f46916efe3730d04fb607c0391."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/robust_lr/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/robust_lr",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 7634 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 9 actual rows from 3 pinned raw source files for source key robust_lr; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "id": "robust_lr__choice__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Preserve source Choice option IDs/descriptions and order; sampled option lists can be instance-specific.",
          "selector": {
            "decision_id": "relation",
            "kind": "judgment",
            "type": "choice"
          }
        }
      ],
      "source_profile": "robust_lr"
    },
    "ruletaker": {
      "disposition": "include",
      "disposition_reason": "Reviewed English typed judgment conversion; preserve selected source splits and label semantics.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/tasksource/ruletaker/blob/a3e0880baeb6ec3d478f4c4d85afe04b21b6cf7f/README.md",
          "what_verified": "Upstream dataset/card identifier tasksource/ruletaker at pinned revision a3e0880baeb6ec3d478f4c4d85afe04b21b6cf7f; license metadata is ['apache-2.0']. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/tasksource/ruletaker/raw/a3e0880baeb6ec3d478f4c4d85afe04b21b6cf7f/README.md",
          "what_verified": "Primary source card was fetched and read at a3e0880baeb6ec3d478f4c4d85afe04b21b6cf7f; card excerpt: # Dataset Card for \"ruletaker\" https://github.com/allenai/ruletaker ``` @inproceedings{ruletaker2020,"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/ruletaker/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 3 source artifacts; 3 referenced local cache paths exist and integrity digests match. Plan source revision a3e0880baeb6ec3d478f4c4d85afe04b21b6cf7f."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/ruletaker/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/ruletaker",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 487460 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 9 actual rows from 3 pinned raw source files for source key ruletaker; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "false": "The hypothesis is false under the stated facts, rules, and closed-world convention: an unprovable positive proposition is false and its negation is true.",
            "true": "The hypothesis is true under the stated facts, rules, and closed-world convention: an unprovable positive proposition is false and its negation is true."
          },
          "id": "ruletaker__noul__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Source-grounded false/true meanings; see source evidence.",
          "selector": {
            "decision_id": "entailed",
            "kind": "judgment",
            "type": "noul"
          }
        }
      ],
      "source_profile": "ruletaker"
    },
    "scicite": {
      "disposition": "include",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching.",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "6b569c1b7045f187cccc94c38af91faa755178ae",
          "sha256": "3c25ecbf2b3044737e2a3fc90e16e9049fc76efac9b07ab87741cb7d251f94b7",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/tasksource/scicite/6b569c1b7045f187cccc94c38af91faa755178ae/README.md?%2Fdatasets%2Ftasksource%2Fscicite%2Fresolve%2F6b569c1b7045f187cccc94c38af91faa755178ae%2FREADME.md=&etag=%22582d5b5961a90a80b3d3dd52e1befa2a9f709a70%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "scicite:default:test:0",
          "local_path": "output/typed_decisions/permissive-v1/cache/72c83a77cd56ddef2e026a0380c474726d339c3028c617ea1d5deca2e36ff951",
          "original_row_match_status": "matched",
          "sha256": "638f0d3c0b1daa763fa356ba81fac503059497803bb452fe32ebbd585d7c26dc",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://huggingface.co/datasets/tasksource/scicite/resolve/01ee26c0a246aa8f043e29541eff7503b2011dd1/default/test/0000.parquet"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/permissive_adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "background": "Background information",
            "method": "Use of a method",
            "result": "Comparison of results"
          },
          "id": "intent",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "Descriptions are copied from the pinned source conversion’s case criteria. Keep IDs aligned; only expand them from source definitions, never from the observed target.",
          "option_semantics": "fixed_source_taxonomy",
          "selector": {
            "decision_id": "intent",
            "type": "choice"
          }
        }
      ],
      "source_profile": "scicite"
    },
    "scitail": {
      "disposition": "include",
      "disposition_reason": "Use the complete SNLI-format release. Negative means unsupported, not contradicted; group by hypothesis. HF card has no license field; author repository and published data card identify Apache-2.0.; For entailment-style Noul rows, false means the queried support/entailment relation is not established; do not relabel all false examples as contradictions.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/allenai/scitail/blob/0cc4353235b289165dfde1c7c5d1be983f99ce44/README.md",
          "what_verified": "Upstream dataset/card identifier allenai/scitail at pinned revision 0cc4353235b289165dfde1c7c5d1be983f99ce44; license metadata is ['apache-2.0']. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/allenai/scitail/raw/0cc4353235b289165dfde1c7c5d1be983f99ce44/README.md",
          "what_verified": "Primary source card was fetched and read at 0cc4353235b289165dfde1c7c5d1be983f99ce44; card excerpt: # Dataset Card for \"scitail\" ## Table of Contents - [Dataset Description](#dataset-description) - [Dataset Summary](#dataset-summary)"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/scitail/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 3 source artifacts; 3 referenced local cache paths exist and integrity digests match. Plan source revision 0cc4353235b289165dfde1c7c5d1be983f99ce44."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/scitail/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/scitail",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 26993 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 9 actual rows from 3 pinned raw source files for source key scitail; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "false": "The premise does not support the hypothesis.",
            "true": "The premise supports the hypothesis."
          },
          "id": "scitail__noul__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Source-grounded false/true meanings; see source evidence.",
          "selector": {
            "decision_id": "entails",
            "kind": "judgment",
            "type": "noul"
          }
        }
      ],
      "source_profile": "scitail"
    },
    "scone": {
      "disposition": "include",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching.",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "505ed270f9e9865c6cfab1e222d804fedd837127",
          "sha256": "0550b7a09be216f587bb5040c6ba4575c4d06e796b959dae41db008043a715ea",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/tasksource/scone/505ed270f9e9865c6cfab1e222d804fedd837127/README.md?%2Fdatasets%2Ftasksource%2Fscone%2Fresolve%2F505ed270f9e9865c6cfab1e222d804fedd837127%2FREADME.md=&etag=%22dc45b2c967b664f30cab6602d12be6860108260a%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "scone:default:test:0",
          "local_path": "output/typed_decisions/permissive-v1/cache/19abedde6548f4095f1256ae0a23c93aa95581ca01fe77827e56a75370454792",
          "original_row_match_status": "matched",
          "sha256": "da2794950fea54157bc30f6de2b941b7756d26afd9be00713844b5106e7b8feb",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://huggingface.co/datasets/tasksource/scone/resolve/ecd23d9ccbf08b679ffd3684f7632bc783284792/default/test/0000.parquet"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/permissive_adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "contradiction": "The premise guarantees that the hypothesis is false.",
            "entailment": "The premise guarantees that the hypothesis is true.",
            "neutral": "The premise does not determine whether the hypothesis is true."
          },
          "id": "relation",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "Descriptions are copied from the pinned source conversion’s case criteria. Keep IDs aligned; only expand them from source definitions, never from the observed target.",
          "option_semantics": "fixed_source_taxonomy",
          "selector": {
            "decision_id": "relation",
            "type": "choice"
          }
        }
      ],
      "source_profile": "scone"
    },
    "sdoh_nli": {
      "disposition": "include_with_caveat",
      "disposition_reason": "See permissive dataset build note; short task-specific instructions, source-specific label checks and grouped splits. Author license overrides missing mirror metadata where documented.; For entailment-style Noul rows, false means the queried support/entailment relation is not established; do not relabel all false examples as contradictions.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/tasksource/SDOH-NLI/blob/75d4cc5f67ecb3f3c44cf5a3fadc3c9579f2f148/README.md",
          "what_verified": "Upstream dataset/card identifier tasksource/SDOH-NLI at pinned revision 75d4cc5f67ecb3f3c44cf5a3fadc3c9579f2f148; license metadata is ['cc-by-4.0']. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/tasksource/SDOH-NLI/raw/75d4cc5f67ecb3f3c44cf5a3fadc3c9579f2f148/README.md",
          "what_verified": "Primary source card was fetched and read at 75d4cc5f67ecb3f3c44cf5a3fadc3c9579f2f148; card excerpt: --- license: cc-by-4.0 task_categories: - text-classification language: - en ---"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/sdoh_nli/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 3 source artifacts; 3 referenced local cache paths exist and integrity digests match. Plan source revision 75d4cc5f67ecb3f3c44cf5a3fadc3c9579f2f148."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/sdoh_nli/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/sdoh_nli",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 28472 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 9 actual rows from 3 pinned raw source files for source key sdoh_nli; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "false": "The note does not support the statement about the person’s circumstances.",
            "true": "The note supports the statement about the person’s circumstances."
          },
          "id": "sdoh_nli__noul__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Source-grounded false/true meanings; see source evidence.",
          "selector": {
            "decision_id": "supported",
            "kind": "judgment",
            "type": "noul"
          }
        }
      ],
      "source_profile": "sdoh_nli"
    },
    "sgd": {
      "disposition": "include",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching.",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "e852981ae34990f4358979625854259302feaa78",
          "sha256": "e16545f163578037e6b7315b8da243f9f9871f4f071a812ab86b73702f124168",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://raw.githubusercontent.com/google-research-datasets/dstc8-schema-guided-dialogue/e852981ae34990f4358979625854259302feaa78/README.md"
        },
        {
          "adapter_replay": "matched",
          "case_id": "sgd:default:test:0",
          "local_path": "output/typed_decisions/additional20-v1/cache/8c96d753d2efe8b492041c2b3900fc98c607b6a23c004e24bfb17735ad7f87e6",
          "original_row_match_status": "matched",
          "sha256": "1fa13b0fe607100bc66ed15e0490df9c52a2cec087ca9733e0f506755eef0ada",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://raw.githubusercontent.com/google-research-datasets/dstc8-schema-guided-dialogue/e852981ae34990f4358979625854259302feaa78/test/dialogues_001.json"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/additional_adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {},
          "id": "intent",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "Options and their descriptions vary by case; use the row’s own option-ID/description pairs. Do not assign a fixed meaning to option IDs across rows.",
          "option_semantics": "per_case_candidates",
          "preserve_case_options": true,
          "selector": {
            "decision_id": "intent",
            "type": "choice"
          }
        }
      ],
      "source_profile": "sgd"
    },
    "snli": {
      "disposition": "include_with_caveat",
      "disposition_reason": "See research note; adapters and full-data quality checks are not implemented by this survey.; For entailment-style Noul rows, false means the queried support/entailment relation is not established; do not relabel all false examples as contradictions.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/stanfordnlp/snli/blob/cdb5c3d5eed6ead6e5a341c8e56e669bb666725b/README.md",
          "what_verified": "Upstream dataset/card identifier stanfordnlp/snli at pinned revision cdb5c3d5eed6ead6e5a341c8e56e669bb666725b; license metadata is ['cc-by-sa-4.0']. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/stanfordnlp/snli/raw/cdb5c3d5eed6ead6e5a341c8e56e669bb666725b/README.md",
          "what_verified": "Primary source card was fetched and read at cdb5c3d5eed6ead6e5a341c8e56e669bb666725b; card excerpt: # Dataset Card for SNLI ## Table of Contents - [Dataset Description](#dataset-description) - [Dataset Summary](#dataset-summary)"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/snli/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 3 source artifacts; 3 referenced local cache paths exist and integrity digests match. Plan source revision cdb5c3d5eed6ead6e5a341c8e56e669bb666725b."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/snli/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/snli",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 568224 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 9 actual rows from 3 pinned raw source files for source key snli; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "id": "snli__choice__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Preserve source Choice option IDs/descriptions and order; sampled option lists can be instance-specific.",
          "selector": {
            "decision_id": "relation",
            "kind": "judgment",
            "type": "choice"
          }
        }
      ],
      "source_profile": "snli"
    },
    "spartqa": {
      "disposition": "include",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching.",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "d29ba456044649c3c2fe29c5ff429c3e6d80a97b",
          "sha256": "c84222b4437cf925db374cc64025c66756f265f75d6c4436802143b284f284b5",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/tasksource/spartqa-mchoice/d29ba456044649c3c2fe29c5ff429c3e6d80a97b/README.md?%2Fdatasets%2Ftasksource%2Fspartqa-mchoice%2Fresolve%2Fd29ba456044649c3c2fe29c5ff429c3e6d80a97b%2FREADME.md=&etag=%2201a06c440c1ffcb6b1d27225484ec2a7ca362901%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "spartqa:default:test:0",
          "local_path": "output/typed_decisions/additional20-v1/cache/1e12cddce57d9d43c076392a444760459e12acc03669fdbe14d4d7c8a5c932c4",
          "original_row_match_status": "matched",
          "sha256": "c0cd6ca3cd1192e680f468b28d6e9aeb338420fa21aa9bd4cacbd7d1cc2d5e53",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://huggingface.co/datasets/tasksource/spartqa-mchoice/resolve/bc5931e32ec8a98f62a81e0c0294b2eaef485ff1/default/test/0000.parquet"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/additional_adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {},
          "id": "answer",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "Options and their descriptions vary by case; use the row’s own option-ID/description pairs. Do not assign a fixed meaning to option IDs across rows.",
          "option_semantics": "per_case_candidates",
          "preserve_case_options": true,
          "selector": {
            "decision_id": "answer",
            "type": "choice"
          }
        }
      ],
      "source_profile": "spartqa"
    },
    "stepgame": {
      "disposition": "include",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching.",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "3ba235d4a7056cf8b062a57aa489f7bc14bff229",
          "sha256": "a00cf76782d1db2d2f3ecd1db4b097d83bbabe4d95fadab8bedb7858009e8266",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/tasksource/stepgame/3ba235d4a7056cf8b062a57aa489f7bc14bff229/README.md?%2Fdatasets%2Ftasksource%2Fstepgame%2Fresolve%2F3ba235d4a7056cf8b062a57aa489f7bc14bff229%2FREADME.md=&etag=%22ffa3754e2c9a5e83867781ee1768df10d130ed0d%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "stepgame:default:test:0",
          "local_path": "output/typed_decisions/permissive-v1/cache/567caee047674ba00fe5403589361a6b43796e8d45742a86d4a2c6e58ee379a3",
          "original_row_match_status": "matched",
          "sha256": "fd53c285bcb8678d146a3e4b622c2de082255cf623f1f9011c5b15dfbbcfc8ea",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://huggingface.co/datasets/tasksource/stepgame/resolve/b1c84cf7a2d8510a38e5d78a74ef2c577df5958b/default/test/0000.parquet"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/permissive_adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "above": "above",
            "below": "below",
            "left": "left",
            "lower-left": "lower left",
            "lower-right": "lower right",
            "overlap": "overlap",
            "right": "right",
            "upper-left": "upper left",
            "upper-right": "upper right"
          },
          "id": "direction",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "Descriptions are copied from the pinned source conversion’s case criteria. Keep IDs aligned; only expand them from source definitions, never from the observed target.",
          "option_semantics": "fixed_source_taxonomy",
          "selector": {
            "decision_id": "direction",
            "type": "choice"
          }
        }
      ],
      "source_profile": "stepgame"
    },
    "synthetic_relevance__nanobeir__NanoArguAna": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Preserve this exact independent teacher Score task and rubric. The assignment reports 25 source-positive scores raised to >=0.75; where the original teacher gave a lower score, this can overrule the answer-evidence rubric. Treat targets as corrected teacher labels, not human gold or official source-qrel metrics; do not normalize across documents.",
      "evidence": [
        {
          "path_or_url": "output/bekko-system-one-dataset-v1/synthetic_relevance__nanobeir__NanoArguAna",
          "what_verified": "Loaded this subset from the active release and sampled {'test': 5} rows: index 0, floor(n/2), and n-1 in each split; split populations are {'test': 400}."
        },
        {
          "path_or_url": "output/bekko-renew-audit/retrieval_samples.jsonl",
          "what_verified": "Contains the exact sampled inference inputs, current instructions, candidates, targets, and provenance at the per-sample line numbers listed in sample_refs."
        },
        {
          "path_or_url": "output/synthetic_dataset/nanobeir-en-normalized-v1/scores.jsonl",
          "what_verified": "Original saved teacher records were directly matched for every sampled row (text, source role, raw score, and corrected target): 5/5. Source-positive floor metadata reports 25 corrected document scores in this subset; raw teacher scores remain available."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/judge_prompt_scores_only.txt",
          "what_verified": "The teacher is explicitly asked for each document’s expected marginal usefulness for producing a correct, specific, well-grounded answer if added alone to RAG context; score 3/4 require explicit answer facts."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/score.py",
          "what_verified": "Confirms the source prompt, input construction (query plus documents), score validation, and saved score outputs; source roles and earlier scores are not shown to the judge."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/normalize_scores.py",
          "what_verified": "Keeps original 0–4 teacher values and scales them by 1/4; attaches the answer-evidence rubric."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/fix_positive_scores.py",
          "what_verified": "Raises each source-positive raw 0/1/2 rating to 3, preserves original and corrected scores, and does not alter source-negative scores."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/hakari-bench/NanoBEIR-en",
          "what_verified": "Original source-card reference in the pinned raw-source catalog: NanoArguAna; checked revision not recorded, declared license unknown."
        },
        {
          "path_or_url": "output/bekko-system-one-dataset-v1/dataset_metadata.json",
          "what_verified": "Subset source-task card states category='natural_language'; A compact ArguAna split where each query is an argument and relevant documents are strong counterarguments. It tests whether retrieval models capture argumentative stance and semantic opposition beyond lexical overlap. This split is the compact NanoArguAna subset used by NanoBEIR-en.. Current subset is test-only and source positive qrels affect positive-floor correction."
        },
        {
          "path_or_url": "output/bekko-renew-audit/retrieval-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/judge_prompt_scores_only.txt",
          "what_verified": "Original teacher rubric uses integer0–4; native criteria values are0–4. Renew aligns instruction numerals to those unchanged values. The old0–1 presentation remains recoverable through value_scale=0.25 in target metadata; targets and rubric meanings are unchanged."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "0": "Not useful or wrong target.",
            "1": "Topic overlap or weak adjacent context.",
            "2": "Useful context about the exact target without the requested answer.",
            "3": "Explicit facts answering an important part for the exact target.",
            "4": "Direct sufficient answer."
          },
          "id": "relevance",
          "instructions": {
            "mode": "replace",
            "value": "Rate the document independently for its usefulness in answering the query, using the 0–4 criterion levels. Use only the query and document. Choose the lowest fitting level. Levels 3 and 4 require explicit answer facts about the exact target; topical overlap alone is insufficient."
          },
          "selector": {
            "decision_id": "relevance",
            "kind": "judgment",
            "type": "score"
          }
        }
      ],
      "source_profile": "synthetic_relevance__nanobeir__NanoArguAna"
    },
    "synthetic_relevance__nanobeir__NanoClimateFEVER": {
      "disposition": "quarantine",
      "disposition_reason": "Quarantine from the renewed answer-evidence Score pool because an inspected source-positive sample directly conflicts with the rubric: The query claims Arctic sea-ice loss is matched by Antarctic sea-ice gain; the source-positive document is an encyclopedia overview of Antarctica that does not state this comparison. The teacher assigned 0, then the qrel-positive floor changed it to 3. Keep the saved raw teacher score and source label separately; do not replace the teacher instruction with source-task semantics while retaining the floored target. Evidence case: synthetic_relevance__nanobeir__NanoClimateFEVER:hakari-bench/NanoBEIR-en/NanoClimateFEVER:1575:doc-1.",
      "evidence": [
        {
          "path_or_url": "output/bekko-system-one-dataset-v1/synthetic_relevance__nanobeir__NanoClimateFEVER",
          "what_verified": "Loaded this subset from the active release and sampled {'test': 4} rows: index 0, floor(n/2), and n-1 in each split; split populations are {'test': 400}."
        },
        {
          "path_or_url": "output/bekko-renew-audit/retrieval_samples.jsonl",
          "what_verified": "Contains the exact sampled inference inputs, current instructions, candidates, targets, and provenance at the per-sample line numbers listed in sample_refs."
        },
        {
          "path_or_url": "output/synthetic_dataset/nanobeir-en-normalized-v1/scores.jsonl",
          "what_verified": "Original saved teacher records were directly matched for every sampled row (text, source role, raw score, and corrected target): 4/4. Source-positive floor metadata reports 68 corrected document scores in this subset; raw teacher scores remain available."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/judge_prompt_scores_only.txt",
          "what_verified": "The teacher is explicitly asked for each document’s expected marginal usefulness for producing a correct, specific, well-grounded answer if added alone to RAG context; score 3/4 require explicit answer facts."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/score.py",
          "what_verified": "Confirms the source prompt, input construction (query plus documents), score validation, and saved score outputs; source roles and earlier scores are not shown to the judge."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/normalize_scores.py",
          "what_verified": "Keeps original 0–4 teacher values and scales them by 1/4; attaches the answer-evidence rubric."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/fix_positive_scores.py",
          "what_verified": "Raises each source-positive raw 0/1/2 rating to 3, preserves original and corrected scores, and does not alter source-negative scores."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/hakari-bench/NanoBEIR-en",
          "what_verified": "Original source-card reference in the pinned raw-source catalog: NanoClimateFEVER; checked revision not recorded, declared license unknown."
        },
        {
          "path_or_url": "output/bekko-system-one-dataset-v1/dataset_metadata.json",
          "what_verified": "Subset source-task card states category='natural_language'; A compact Climate-FEVER split for retrieving evidence documents for real-world climate claims. It measures claim-evidence matching in a fact-checking setting with climate-domain terminology. This split is the compact NanoClimateFEVER subset used by NanoBEIR-en.. Current subset is test-only and source positive qrels affect positive-floor correction."
        },
        {
          "path_or_url": "output/bekko-renew-audit/retrieval-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/judge_prompt_scores_only.txt",
          "what_verified": "Original teacher rubric uses integer0–4; native criteria values are0–4. Renew aligns instruction numerals to those unchanged values. The old0–1 presentation remains recoverable through value_scale=0.25 in target metadata; targets and rubric meanings are unchanged."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "0": "Not useful or wrong target.",
            "1": "Topic overlap or weak adjacent context.",
            "2": "Useful context about the exact target without the requested answer.",
            "3": "Explicit facts answering an important part for the exact target.",
            "4": "Direct sufficient answer."
          },
          "id": "relevance",
          "instructions": {
            "mode": "replace",
            "value": "Rate the document independently for its usefulness in answering the query, using the 0–4 criterion levels. Use only the query and document. Choose the lowest fitting level. Levels 3 and 4 require explicit answer facts about the exact target; topical overlap alone is insufficient."
          },
          "selector": {
            "decision_id": "relevance",
            "kind": "judgment",
            "type": "score"
          }
        }
      ],
      "source_profile": "synthetic_relevance__nanobeir__NanoClimateFEVER"
    },
    "synthetic_relevance__nanobeir__NanoDBPedia": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Preserve this exact independent teacher Score task and rubric. The assignment reports 67 source-positive scores raised to >=0.75; where the original teacher gave a lower score, this can overrule the answer-evidence rubric. Treat targets as corrected teacher labels, not human gold or official source-qrel metrics; do not normalize across documents.",
      "evidence": [
        {
          "path_or_url": "output/bekko-system-one-dataset-v1/synthetic_relevance__nanobeir__NanoDBPedia",
          "what_verified": "Loaded this subset from the active release and sampled {'test': 5} rows: index 0, floor(n/2), and n-1 in each split; split populations are {'test': 400}."
        },
        {
          "path_or_url": "output/bekko-renew-audit/retrieval_samples.jsonl",
          "what_verified": "Contains the exact sampled inference inputs, current instructions, candidates, targets, and provenance at the per-sample line numbers listed in sample_refs."
        },
        {
          "path_or_url": "output/synthetic_dataset/nanobeir-en-normalized-v1/scores.jsonl",
          "what_verified": "Original saved teacher records were directly matched for every sampled row (text, source role, raw score, and corrected target): 5/5. Source-positive floor metadata reports 67 corrected document scores in this subset; raw teacher scores remain available."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/judge_prompt_scores_only.txt",
          "what_verified": "The teacher is explicitly asked for each document’s expected marginal usefulness for producing a correct, specific, well-grounded answer if added alone to RAG context; score 3/4 require explicit answer facts."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/score.py",
          "what_verified": "Confirms the source prompt, input construction (query plus documents), score validation, and saved score outputs; source roles and earlier scores are not shown to the judge."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/normalize_scores.py",
          "what_verified": "Keeps original 0–4 teacher values and scales them by 1/4; attaches the answer-evidence rubric."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/fix_positive_scores.py",
          "what_verified": "Raises each source-positive raw 0/1/2 rating to 3, preserves original and corrected scores, and does not alter source-negative scores."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/hakari-bench/NanoBEIR-en",
          "what_verified": "Original source-card reference in the pinned raw-source catalog: NanoDBPedia; checked revision not recorded, declared license unknown."
        },
        {
          "path_or_url": "output/bekko-system-one-dataset-v1/dataset_metadata.json",
          "what_verified": "Subset source-task card states category='natural_language'; A compact DBpedia entity-search split where natural-language queries must retrieve relevant DBpedia entity documents. It emphasizes entity disambiguation and matching descriptions to knowledge-base pages. This split is the compact NanoDBPedia subset used by NanoBEIR-en.. Current subset is test-only and source positive qrels affect positive-floor correction."
        },
        {
          "path_or_url": "output/bekko-renew-audit/retrieval-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/judge_prompt_scores_only.txt",
          "what_verified": "Original teacher rubric uses integer0–4; native criteria values are0–4. Renew aligns instruction numerals to those unchanged values. The old0–1 presentation remains recoverable through value_scale=0.25 in target metadata; targets and rubric meanings are unchanged."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "0": "Not useful or wrong target.",
            "1": "Topic overlap or weak adjacent context.",
            "2": "Useful context about the exact target without the requested answer.",
            "3": "Explicit facts answering an important part for the exact target.",
            "4": "Direct sufficient answer."
          },
          "id": "relevance",
          "instructions": {
            "mode": "replace",
            "value": "Rate the document independently for its usefulness in answering the query, using the 0–4 criterion levels. Use only the query and document. Choose the lowest fitting level. Levels 3 and 4 require explicit answer facts about the exact target; topical overlap alone is insufficient."
          },
          "selector": {
            "decision_id": "relevance",
            "kind": "judgment",
            "type": "score"
          }
        }
      ],
      "source_profile": "synthetic_relevance__nanobeir__NanoDBPedia"
    },
    "synthetic_relevance__nanobeir__NanoFEVER": {
      "disposition": "quarantine",
      "disposition_reason": "Pending task/rubric compatibility review; this does not establish that the original paraphrase label or fact-verification qrel is wrong. Quarantine from the renewed answer-evidence Score pool because an inspected source-positive sample raises an unresolved interpretation conflict with the answer-evidence rubric: The claim says Wolfgang Amadeus Mozart was a wrestler; the source-positive document is a Mozart biography with no wrestling evidence. The teacher assigned 0, then the qrel-positive floor changed it to 3. Keep the saved raw teacher score and source label separately; do not replace the teacher instruction with source-task semantics while retaining the floored target. Evidence case: synthetic_relevance__nanobeir__NanoFEVER:hakari-bench/NanoBEIR-en/NanoFEVER:34583:doc-8.",
      "evidence": [
        {
          "path_or_url": "output/bekko-system-one-dataset-v1/synthetic_relevance__nanobeir__NanoFEVER",
          "what_verified": "Loaded this subset from the active release and sampled {'test': 5} rows: index 0, floor(n/2), and n-1 in each split; split populations are {'test': 400}."
        },
        {
          "path_or_url": "output/bekko-renew-audit/retrieval_samples.jsonl",
          "what_verified": "Contains the exact sampled inference inputs, current instructions, candidates, targets, and provenance at the per-sample line numbers listed in sample_refs."
        },
        {
          "path_or_url": "output/synthetic_dataset/nanobeir-en-normalized-v1/scores.jsonl",
          "what_verified": "Original saved teacher records were directly matched for every sampled row (text, source role, raw score, and corrected target): 5/5. Source-positive floor metadata reports 14 corrected document scores in this subset; raw teacher scores remain available."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/judge_prompt_scores_only.txt",
          "what_verified": "The teacher is explicitly asked for each document’s expected marginal usefulness for producing a correct, specific, well-grounded answer if added alone to RAG context; score 3/4 require explicit answer facts."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/score.py",
          "what_verified": "Confirms the source prompt, input construction (query plus documents), score validation, and saved score outputs; source roles and earlier scores are not shown to the judge."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/normalize_scores.py",
          "what_verified": "Keeps original 0–4 teacher values and scales them by 1/4; attaches the answer-evidence rubric."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/fix_positive_scores.py",
          "what_verified": "Raises each source-positive raw 0/1/2 rating to 3, preserves original and corrected scores, and does not alter source-negative scores."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/hakari-bench/NanoBEIR-en",
          "what_verified": "Original source-card reference in the pinned raw-source catalog: NanoFEVER; checked revision not recorded, declared license unknown."
        },
        {
          "path_or_url": "output/bekko-system-one-dataset-v1/dataset_metadata.json",
          "what_verified": "Subset source-task card states category='natural_language'; A compact FEVER split for retrieving Wikipedia evidence relevant to factual claims. It evaluates fact-checking retrieval where claims must be matched to evidence-bearing documents. This split is the compact NanoFEVER subset used by NanoBEIR-en.. Current subset is test-only and source positive qrels affect positive-floor correction."
        },
        {
          "path_or_url": "output/bekko-renew-audit/retrieval-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/judge_prompt_scores_only.txt",
          "what_verified": "Original teacher rubric uses integer0–4; native criteria values are0–4. Renew aligns instruction numerals to those unchanged values. The old0–1 presentation remains recoverable through value_scale=0.25 in target metadata; targets and rubric meanings are unchanged."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "0": "Not useful or wrong target.",
            "1": "Topic overlap or weak adjacent context.",
            "2": "Useful context about the exact target without the requested answer.",
            "3": "Explicit facts answering an important part for the exact target.",
            "4": "Direct sufficient answer."
          },
          "id": "relevance",
          "instructions": {
            "mode": "replace",
            "value": "Rate the document independently for its usefulness in answering the query, using the 0–4 criterion levels. Use only the query and document. Choose the lowest fitting level. Levels 3 and 4 require explicit answer facts about the exact target; topical overlap alone is insufficient."
          },
          "selector": {
            "decision_id": "relevance",
            "kind": "judgment",
            "type": "score"
          }
        }
      ],
      "source_profile": "synthetic_relevance__nanobeir__NanoFEVER"
    },
    "synthetic_relevance__nanobeir__NanoFiQA2018": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Preserve this exact independent teacher Score task and rubric. The assignment reports 21 source-positive scores raised to >=0.75; where the original teacher gave a lower score, this can overrule the answer-evidence rubric. Treat targets as corrected teacher labels, not human gold or official source-qrel metrics; do not normalize across documents.",
      "evidence": [
        {
          "path_or_url": "output/bekko-system-one-dataset-v1/synthetic_relevance__nanobeir__NanoFiQA2018",
          "what_verified": "Loaded this subset from the active release and sampled {'test': 5} rows: index 0, floor(n/2), and n-1 in each split; split populations are {'test': 400}."
        },
        {
          "path_or_url": "output/bekko-renew-audit/retrieval_samples.jsonl",
          "what_verified": "Contains the exact sampled inference inputs, current instructions, candidates, targets, and provenance at the per-sample line numbers listed in sample_refs."
        },
        {
          "path_or_url": "output/synthetic_dataset/nanobeir-en-normalized-v1/scores.jsonl",
          "what_verified": "Original saved teacher records were directly matched for every sampled row (text, source role, raw score, and corrected target): 5/5. Source-positive floor metadata reports 21 corrected document scores in this subset; raw teacher scores remain available."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/judge_prompt_scores_only.txt",
          "what_verified": "The teacher is explicitly asked for each document’s expected marginal usefulness for producing a correct, specific, well-grounded answer if added alone to RAG context; score 3/4 require explicit answer facts."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/score.py",
          "what_verified": "Confirms the source prompt, input construction (query plus documents), score validation, and saved score outputs; source roles and earlier scores are not shown to the judge."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/normalize_scores.py",
          "what_verified": "Keeps original 0–4 teacher values and scales them by 1/4; attaches the answer-evidence rubric."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/fix_positive_scores.py",
          "what_verified": "Raises each source-positive raw 0/1/2 rating to 3, preserves original and corrected scores, and does not alter source-negative scores."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/hakari-bench/NanoBEIR-en",
          "what_verified": "Original source-card reference in the pinned raw-source catalog: NanoFiQA2018; checked revision not recorded, declared license unknown."
        },
        {
          "path_or_url": "output/bekko-system-one-dataset-v1/dataset_metadata.json",
          "what_verified": "Subset source-task card states category='natural_language'; A compact FiQA-2018 retrieval split focused on finance-domain questions and opinion-bearing answers. It tests retrieval under specialized financial terminology and user question phrasing. This split is the compact NanoFiQA2018 subset used by NanoBEIR-en.. Current subset is test-only and source positive qrels affect positive-floor correction."
        },
        {
          "path_or_url": "output/bekko-renew-audit/retrieval-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/judge_prompt_scores_only.txt",
          "what_verified": "Original teacher rubric uses integer0–4; native criteria values are0–4. Renew aligns instruction numerals to those unchanged values. The old0–1 presentation remains recoverable through value_scale=0.25 in target metadata; targets and rubric meanings are unchanged."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "0": "Not useful or wrong target.",
            "1": "Topic overlap or weak adjacent context.",
            "2": "Useful context about the exact target without the requested answer.",
            "3": "Explicit facts answering an important part for the exact target.",
            "4": "Direct sufficient answer."
          },
          "id": "relevance",
          "instructions": {
            "mode": "replace",
            "value": "Rate the document independently for its usefulness in answering the query, using the 0–4 criterion levels. Use only the query and document. Choose the lowest fitting level. Levels 3 and 4 require explicit answer facts about the exact target; topical overlap alone is insufficient."
          },
          "selector": {
            "decision_id": "relevance",
            "kind": "judgment",
            "type": "score"
          }
        }
      ],
      "source_profile": "synthetic_relevance__nanobeir__NanoFiQA2018"
    },
    "synthetic_relevance__nanobeir__NanoHotpotQA": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Preserve this exact independent teacher Score task and rubric. The assignment reports 28 source-positive scores raised to >=0.75; where the original teacher gave a lower score, this can overrule the answer-evidence rubric. Treat targets as corrected teacher labels, not human gold or official source-qrel metrics; do not normalize across documents.",
      "evidence": [
        {
          "path_or_url": "output/bekko-system-one-dataset-v1/synthetic_relevance__nanobeir__NanoHotpotQA",
          "what_verified": "Loaded this subset from the active release and sampled {'test': 5} rows: index 0, floor(n/2), and n-1 in each split; split populations are {'test': 400}."
        },
        {
          "path_or_url": "output/bekko-renew-audit/retrieval_samples.jsonl",
          "what_verified": "Contains the exact sampled inference inputs, current instructions, candidates, targets, and provenance at the per-sample line numbers listed in sample_refs."
        },
        {
          "path_or_url": "output/synthetic_dataset/nanobeir-en-normalized-v1/scores.jsonl",
          "what_verified": "Original saved teacher records were directly matched for every sampled row (text, source role, raw score, and corrected target): 5/5. Source-positive floor metadata reports 28 corrected document scores in this subset; raw teacher scores remain available."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/judge_prompt_scores_only.txt",
          "what_verified": "The teacher is explicitly asked for each document’s expected marginal usefulness for producing a correct, specific, well-grounded answer if added alone to RAG context; score 3/4 require explicit answer facts."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/score.py",
          "what_verified": "Confirms the source prompt, input construction (query plus documents), score validation, and saved score outputs; source roles and earlier scores are not shown to the judge."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/normalize_scores.py",
          "what_verified": "Keeps original 0–4 teacher values and scales them by 1/4; attaches the answer-evidence rubric."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/fix_positive_scores.py",
          "what_verified": "Raises each source-positive raw 0/1/2 rating to 3, preserves original and corrected scores, and does not alter source-negative scores."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/hakari-bench/NanoBEIR-en",
          "what_verified": "Original source-card reference in the pinned raw-source catalog: NanoHotpotQA; checked revision not recorded, declared license unknown."
        },
        {
          "path_or_url": "output/bekko-system-one-dataset-v1/dataset_metadata.json",
          "what_verified": "Subset source-task card states category='natural_language'; A compact HotpotQA retrieval split where questions require finding relevant Wikipedia evidence. It emphasizes multi-hop question intent and paragraph-level evidence retrieval. This split is the compact NanoHotpotQA subset used by NanoBEIR-en.. Current subset is test-only and source positive qrels affect positive-floor correction."
        },
        {
          "path_or_url": "output/bekko-renew-audit/retrieval-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/judge_prompt_scores_only.txt",
          "what_verified": "Original teacher rubric uses integer0–4; native criteria values are0–4. Renew aligns instruction numerals to those unchanged values. The old0–1 presentation remains recoverable through value_scale=0.25 in target metadata; targets and rubric meanings are unchanged."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "0": "Not useful or wrong target.",
            "1": "Topic overlap or weak adjacent context.",
            "2": "Useful context about the exact target without the requested answer.",
            "3": "Explicit facts answering an important part for the exact target.",
            "4": "Direct sufficient answer."
          },
          "id": "relevance",
          "instructions": {
            "mode": "replace",
            "value": "Rate the document independently for its usefulness in answering the query, using the 0–4 criterion levels. Use only the query and document. Choose the lowest fitting level. Levels 3 and 4 require explicit answer facts about the exact target; topical overlap alone is insufficient."
          },
          "selector": {
            "decision_id": "relevance",
            "kind": "judgment",
            "type": "score"
          }
        }
      ],
      "source_profile": "synthetic_relevance__nanobeir__NanoHotpotQA"
    },
    "synthetic_relevance__nanobeir__NanoMSMARCO": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Preserve this exact independent teacher Score task and rubric. The assignment reports 12 source-positive scores raised to >=0.75; where the original teacher gave a lower score, this can overrule the answer-evidence rubric. Treat targets as corrected teacher labels, not human gold or official source-qrel metrics; do not normalize across documents.",
      "evidence": [
        {
          "path_or_url": "output/bekko-system-one-dataset-v1/synthetic_relevance__nanobeir__NanoMSMARCO",
          "what_verified": "Loaded this subset from the active release and sampled {'test': 5} rows: index 0, floor(n/2), and n-1 in each split; split populations are {'test': 400}."
        },
        {
          "path_or_url": "output/bekko-renew-audit/retrieval_samples.jsonl",
          "what_verified": "Contains the exact sampled inference inputs, current instructions, candidates, targets, and provenance at the per-sample line numbers listed in sample_refs."
        },
        {
          "path_or_url": "output/synthetic_dataset/nanobeir-en-normalized-v1/scores.jsonl",
          "what_verified": "Original saved teacher records were directly matched for every sampled row (text, source role, raw score, and corrected target): 5/5. Source-positive floor metadata reports 12 corrected document scores in this subset; raw teacher scores remain available."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/judge_prompt_scores_only.txt",
          "what_verified": "The teacher is explicitly asked for each document’s expected marginal usefulness for producing a correct, specific, well-grounded answer if added alone to RAG context; score 3/4 require explicit answer facts."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/score.py",
          "what_verified": "Confirms the source prompt, input construction (query plus documents), score validation, and saved score outputs; source roles and earlier scores are not shown to the judge."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/normalize_scores.py",
          "what_verified": "Keeps original 0–4 teacher values and scales them by 1/4; attaches the answer-evidence rubric."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/fix_positive_scores.py",
          "what_verified": "Raises each source-positive raw 0/1/2 rating to 3, preserves original and corrected scores, and does not alter source-negative scores."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/hakari-bench/NanoBEIR-en",
          "what_verified": "Original source-card reference in the pinned raw-source catalog: NanoMSMARCO; checked revision not recorded, declared license unknown."
        },
        {
          "path_or_url": "output/bekko-system-one-dataset-v1/dataset_metadata.json",
          "what_verified": "Subset source-task card states category='natural_language'; A compact MS MARCO passage-retrieval split using real web-search style questions. It evaluates matching short user queries to answer-bearing passages. This split is the compact NanoMSMARCO subset used by NanoBEIR-en.. Current subset is test-only and source positive qrels affect positive-floor correction."
        },
        {
          "path_or_url": "output/bekko-renew-audit/retrieval-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/judge_prompt_scores_only.txt",
          "what_verified": "Original teacher rubric uses integer0–4; native criteria values are0–4. Renew aligns instruction numerals to those unchanged values. The old0–1 presentation remains recoverable through value_scale=0.25 in target metadata; targets and rubric meanings are unchanged."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "0": "Not useful or wrong target.",
            "1": "Topic overlap or weak adjacent context.",
            "2": "Useful context about the exact target without the requested answer.",
            "3": "Explicit facts answering an important part for the exact target.",
            "4": "Direct sufficient answer."
          },
          "id": "relevance",
          "instructions": {
            "mode": "replace",
            "value": "Rate the document independently for its usefulness in answering the query, using the 0–4 criterion levels. Use only the query and document. Choose the lowest fitting level. Levels 3 and 4 require explicit answer facts about the exact target; topical overlap alone is insufficient."
          },
          "selector": {
            "decision_id": "relevance",
            "kind": "judgment",
            "type": "score"
          }
        }
      ],
      "source_profile": "synthetic_relevance__nanobeir__NanoMSMARCO"
    },
    "synthetic_relevance__nanobeir__NanoNFCorpus": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Preserve this exact independent teacher Score task and rubric. The assignment reports 72 source-positive scores raised to >=0.75; where the original teacher gave a lower score, this can overrule the answer-evidence rubric. Treat targets as corrected teacher labels, not human gold or official source-qrel metrics; do not normalize across documents.",
      "evidence": [
        {
          "path_or_url": "output/bekko-system-one-dataset-v1/synthetic_relevance__nanobeir__NanoNFCorpus",
          "what_verified": "Loaded this subset from the active release and sampled {'test': 5} rows: index 0, floor(n/2), and n-1 in each split; split populations are {'test': 400}."
        },
        {
          "path_or_url": "output/bekko-renew-audit/retrieval_samples.jsonl",
          "what_verified": "Contains the exact sampled inference inputs, current instructions, candidates, targets, and provenance at the per-sample line numbers listed in sample_refs."
        },
        {
          "path_or_url": "output/synthetic_dataset/nanobeir-en-normalized-v1/scores.jsonl",
          "what_verified": "Original saved teacher records were directly matched for every sampled row (text, source role, raw score, and corrected target): 5/5. Source-positive floor metadata reports 72 corrected document scores in this subset; raw teacher scores remain available."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/judge_prompt_scores_only.txt",
          "what_verified": "The teacher is explicitly asked for each document’s expected marginal usefulness for producing a correct, specific, well-grounded answer if added alone to RAG context; score 3/4 require explicit answer facts."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/score.py",
          "what_verified": "Confirms the source prompt, input construction (query plus documents), score validation, and saved score outputs; source roles and earlier scores are not shown to the judge."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/normalize_scores.py",
          "what_verified": "Keeps original 0–4 teacher values and scales them by 1/4; attaches the answer-evidence rubric."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/fix_positive_scores.py",
          "what_verified": "Raises each source-positive raw 0/1/2 rating to 3, preserves original and corrected scores, and does not alter source-negative scores."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/hakari-bench/NanoBEIR-en",
          "what_verified": "Original source-card reference in the pinned raw-source catalog: NanoNFCorpus; checked revision not recorded, declared license unknown."
        },
        {
          "path_or_url": "output/bekko-system-one-dataset-v1/dataset_metadata.json",
          "what_verified": "Subset source-task card states category='natural_language'; A compact NFCorpus split for medical information retrieval. Queries are natural-language health information needs, and relevant documents come from medically oriented full-text content. This split is the compact NanoNFCorpus subset used by NanoBEIR-en.. Current subset is test-only and source positive qrels affect positive-floor correction."
        },
        {
          "path_or_url": "output/bekko-renew-audit/retrieval-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/judge_prompt_scores_only.txt",
          "what_verified": "Original teacher rubric uses integer0–4; native criteria values are0–4. Renew aligns instruction numerals to those unchanged values. The old0–1 presentation remains recoverable through value_scale=0.25 in target metadata; targets and rubric meanings are unchanged."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "0": "Not useful or wrong target.",
            "1": "Topic overlap or weak adjacent context.",
            "2": "Useful context about the exact target without the requested answer.",
            "3": "Explicit facts answering an important part for the exact target.",
            "4": "Direct sufficient answer."
          },
          "id": "relevance",
          "instructions": {
            "mode": "replace",
            "value": "Rate the document independently for its usefulness in answering the query, using the 0–4 criterion levels. Use only the query and document. Choose the lowest fitting level. Levels 3 and 4 require explicit answer facts about the exact target; topical overlap alone is insufficient."
          },
          "selector": {
            "decision_id": "relevance",
            "kind": "judgment",
            "type": "score"
          }
        }
      ],
      "source_profile": "synthetic_relevance__nanobeir__NanoNFCorpus"
    },
    "synthetic_relevance__nanobeir__NanoNQ": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Preserve this exact independent teacher Score task and rubric. The assignment reports 9 source-positive scores raised to >=0.75; where the original teacher gave a lower score, this can overrule the answer-evidence rubric. Treat targets as corrected teacher labels, not human gold or official source-qrel metrics; do not normalize across documents.",
      "evidence": [
        {
          "path_or_url": "output/bekko-system-one-dataset-v1/synthetic_relevance__nanobeir__NanoNQ",
          "what_verified": "Loaded this subset from the active release and sampled {'test': 5} rows: index 0, floor(n/2), and n-1 in each split; split populations are {'test': 400}."
        },
        {
          "path_or_url": "output/bekko-renew-audit/retrieval_samples.jsonl",
          "what_verified": "Contains the exact sampled inference inputs, current instructions, candidates, targets, and provenance at the per-sample line numbers listed in sample_refs."
        },
        {
          "path_or_url": "output/synthetic_dataset/nanobeir-en-normalized-v1/scores.jsonl",
          "what_verified": "Original saved teacher records were directly matched for every sampled row (text, source role, raw score, and corrected target): 5/5. Source-positive floor metadata reports 9 corrected document scores in this subset; raw teacher scores remain available."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/judge_prompt_scores_only.txt",
          "what_verified": "The teacher is explicitly asked for each document’s expected marginal usefulness for producing a correct, specific, well-grounded answer if added alone to RAG context; score 3/4 require explicit answer facts."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/score.py",
          "what_verified": "Confirms the source prompt, input construction (query plus documents), score validation, and saved score outputs; source roles and earlier scores are not shown to the judge."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/normalize_scores.py",
          "what_verified": "Keeps original 0–4 teacher values and scales them by 1/4; attaches the answer-evidence rubric."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/fix_positive_scores.py",
          "what_verified": "Raises each source-positive raw 0/1/2 rating to 3, preserves original and corrected scores, and does not alter source-negative scores."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/hakari-bench/NanoBEIR-en",
          "what_verified": "Original source-card reference in the pinned raw-source catalog: NanoNQ; checked revision not recorded, declared license unknown."
        },
        {
          "path_or_url": "output/bekko-system-one-dataset-v1/dataset_metadata.json",
          "what_verified": "Subset source-task card states category='natural_language'; A compact Natural Questions retrieval split built from real user questions. It tests whether models retrieve Wikipedia documents or passages that contain answers to information-seeking questions. This split is the compact NanoNQ subset used by NanoBEIR-en.. Current subset is test-only and source positive qrels affect positive-floor correction."
        },
        {
          "path_or_url": "output/bekko-renew-audit/retrieval-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/judge_prompt_scores_only.txt",
          "what_verified": "Original teacher rubric uses integer0–4; native criteria values are0–4. Renew aligns instruction numerals to those unchanged values. The old0–1 presentation remains recoverable through value_scale=0.25 in target metadata; targets and rubric meanings are unchanged."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "0": "Not useful or wrong target.",
            "1": "Topic overlap or weak adjacent context.",
            "2": "Useful context about the exact target without the requested answer.",
            "3": "Explicit facts answering an important part for the exact target.",
            "4": "Direct sufficient answer."
          },
          "id": "relevance",
          "instructions": {
            "mode": "replace",
            "value": "Rate the document independently for its usefulness in answering the query, using the 0–4 criterion levels. Use only the query and document. Choose the lowest fitting level. Levels 3 and 4 require explicit answer facts about the exact target; topical overlap alone is insufficient."
          },
          "selector": {
            "decision_id": "relevance",
            "kind": "judgment",
            "type": "score"
          }
        }
      ],
      "source_profile": "synthetic_relevance__nanobeir__NanoNQ"
    },
    "synthetic_relevance__nanobeir__NanoQuoraRetrieval": {
      "disposition": "quarantine",
      "disposition_reason": "Quarantine from the renewed answer-evidence Score pool because an inspected source-positive sample directly conflicts with the rubric: The query asks what “being civilized” means; the source-positive duplicate question asks the same thing but contains no definition or answer. The teacher assigned 0, then the qrel-positive floor changed it to 3. Keep the saved raw teacher score and source label separately; do not replace the teacher instruction with source-task semantics while retaining the floored target. Evidence case: synthetic_relevance__nanobeir__NanoQuoraRetrieval:hakari-bench/NanoBEIR-en/NanoQuoraRetrieval:56198:doc-1.",
      "evidence": [
        {
          "path_or_url": "output/bekko-system-one-dataset-v1/synthetic_relevance__nanobeir__NanoQuoraRetrieval",
          "what_verified": "Loaded this subset from the active release and sampled {'test': 5} rows: index 0, floor(n/2), and n-1 in each split; split populations are {'test': 400}."
        },
        {
          "path_or_url": "output/bekko-renew-audit/retrieval_samples.jsonl",
          "what_verified": "Contains the exact sampled inference inputs, current instructions, candidates, targets, and provenance at the per-sample line numbers listed in sample_refs."
        },
        {
          "path_or_url": "output/synthetic_dataset/nanobeir-en-normalized-v1/scores.jsonl",
          "what_verified": "Original saved teacher records were directly matched for every sampled row (text, source role, raw score, and corrected target): 5/5. Source-positive floor metadata reports 26 corrected document scores in this subset; raw teacher scores remain available."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/judge_prompt_scores_only.txt",
          "what_verified": "The teacher is explicitly asked for each document’s expected marginal usefulness for producing a correct, specific, well-grounded answer if added alone to RAG context; score 3/4 require explicit answer facts."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/score.py",
          "what_verified": "Confirms the source prompt, input construction (query plus documents), score validation, and saved score outputs; source roles and earlier scores are not shown to the judge."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/normalize_scores.py",
          "what_verified": "Keeps original 0–4 teacher values and scales them by 1/4; attaches the answer-evidence rubric."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/fix_positive_scores.py",
          "what_verified": "Raises each source-positive raw 0/1/2 rating to 3, preserves original and corrected scores, and does not alter source-negative scores."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/hakari-bench/NanoBEIR-en",
          "what_verified": "Original source-card reference in the pinned raw-source catalog: NanoQuoraRetrieval; checked revision not recorded, declared license unknown."
        },
        {
          "path_or_url": "output/bekko-system-one-dataset-v1/dataset_metadata.json",
          "what_verified": "Subset source-task card states category='natural_language'; A compact Quora duplicate-question retrieval split. Queries are questions and relevant documents are semantically duplicate questions, testing paraphrase-level matching. This split is the compact NanoQuoraRetrieval subset used by NanoBEIR-en.. Current subset is test-only and source positive qrels affect positive-floor correction."
        },
        {
          "path_or_url": "output/bekko-renew-audit/retrieval-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/judge_prompt_scores_only.txt",
          "what_verified": "Original teacher rubric uses integer0–4; native criteria values are0–4. Renew aligns instruction numerals to those unchanged values. The old0–1 presentation remains recoverable through value_scale=0.25 in target metadata; targets and rubric meanings are unchanged."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "0": "Not useful or wrong target.",
            "1": "Topic overlap or weak adjacent context.",
            "2": "Useful context about the exact target without the requested answer.",
            "3": "Explicit facts answering an important part for the exact target.",
            "4": "Direct sufficient answer."
          },
          "id": "relevance",
          "instructions": {
            "mode": "replace",
            "value": "Rate the document independently for its usefulness in answering the query, using the 0–4 criterion levels. Use only the query and document. Choose the lowest fitting level. Levels 3 and 4 require explicit answer facts about the exact target; topical overlap alone is insufficient."
          },
          "selector": {
            "decision_id": "relevance",
            "kind": "judgment",
            "type": "score"
          }
        }
      ],
      "source_profile": "synthetic_relevance__nanobeir__NanoQuoraRetrieval"
    },
    "synthetic_relevance__nanobeir__NanoSCIDOCS": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Preserve this exact independent teacher Score task and rubric. The assignment reports 75 source-positive scores raised to >=0.75; where the original teacher gave a lower score, this can overrule the answer-evidence rubric. Treat targets as corrected teacher labels, not human gold or official source-qrel metrics; do not normalize across documents.",
      "evidence": [
        {
          "path_or_url": "output/bekko-system-one-dataset-v1/synthetic_relevance__nanobeir__NanoSCIDOCS",
          "what_verified": "Loaded this subset from the active release and sampled {'test': 5} rows: index 0, floor(n/2), and n-1 in each split; split populations are {'test': 400}."
        },
        {
          "path_or_url": "output/bekko-renew-audit/retrieval_samples.jsonl",
          "what_verified": "Contains the exact sampled inference inputs, current instructions, candidates, targets, and provenance at the per-sample line numbers listed in sample_refs."
        },
        {
          "path_or_url": "output/synthetic_dataset/nanobeir-en-normalized-v1/scores.jsonl",
          "what_verified": "Original saved teacher records were directly matched for every sampled row (text, source role, raw score, and corrected target): 5/5. Source-positive floor metadata reports 75 corrected document scores in this subset; raw teacher scores remain available."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/judge_prompt_scores_only.txt",
          "what_verified": "The teacher is explicitly asked for each document’s expected marginal usefulness for producing a correct, specific, well-grounded answer if added alone to RAG context; score 3/4 require explicit answer facts."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/score.py",
          "what_verified": "Confirms the source prompt, input construction (query plus documents), score validation, and saved score outputs; source roles and earlier scores are not shown to the judge."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/normalize_scores.py",
          "what_verified": "Keeps original 0–4 teacher values and scales them by 1/4; attaches the answer-evidence rubric."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/fix_positive_scores.py",
          "what_verified": "Raises each source-positive raw 0/1/2 rating to 3, preserves original and corrected scores, and does not alter source-negative scores."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/hakari-bench/NanoBEIR-en",
          "what_verified": "Original source-card reference in the pinned raw-source catalog: NanoSCIDOCS; checked revision not recorded, declared license unknown."
        },
        {
          "path_or_url": "output/bekko-system-one-dataset-v1/dataset_metadata.json",
          "what_verified": "Subset source-task card states category='natural_language'; A compact SCIDOCS retrieval split over scientific papers. It evaluates retrieval of related academic documents using citation-informed scholarly relevance signals. This split is the compact NanoSCIDOCS subset used by NanoBEIR-en.. Current subset is test-only and source positive qrels affect positive-floor correction."
        },
        {
          "path_or_url": "output/bekko-renew-audit/retrieval-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/judge_prompt_scores_only.txt",
          "what_verified": "Original teacher rubric uses integer0–4; native criteria values are0–4. Renew aligns instruction numerals to those unchanged values. The old0–1 presentation remains recoverable through value_scale=0.25 in target metadata; targets and rubric meanings are unchanged."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "0": "Not useful or wrong target.",
            "1": "Topic overlap or weak adjacent context.",
            "2": "Useful context about the exact target without the requested answer.",
            "3": "Explicit facts answering an important part for the exact target.",
            "4": "Direct sufficient answer."
          },
          "id": "relevance",
          "instructions": {
            "mode": "replace",
            "value": "Rate the document independently for its usefulness in answering the query, using the 0–4 criterion levels. Use only the query and document. Choose the lowest fitting level. Levels 3 and 4 require explicit answer facts about the exact target; topical overlap alone is insufficient."
          },
          "selector": {
            "decision_id": "relevance",
            "kind": "judgment",
            "type": "score"
          }
        }
      ],
      "source_profile": "synthetic_relevance__nanobeir__NanoSCIDOCS"
    },
    "synthetic_relevance__nanobeir__NanoSciFact": {
      "disposition": "quarantine",
      "disposition_reason": "Quarantine from the renewed answer-evidence Score pool because an inspected source-positive sample directly conflicts with the rubric: The query claims insomnia can be treated with cognitive behavioral therapy; the source-positive document describes circadian-rhythm sleep-disorder symptoms, not CBT treatment efficacy. The teacher assigned 0, then the qrel-positive floor changed it to 3. Keep the saved raw teacher score and source label separately; do not replace the teacher instruction with source-task semantics while retaining the floored target. Evidence case: synthetic_relevance__nanobeir__NanoSciFact:hakari-bench/NanoBEIR-en/NanoSciFact:641:doc-1.",
      "evidence": [
        {
          "path_or_url": "output/bekko-system-one-dataset-v1/synthetic_relevance__nanobeir__NanoSciFact",
          "what_verified": "Loaded this subset from the active release and sampled {'test': 5} rows: index 0, floor(n/2), and n-1 in each split; split populations are {'test': 400}."
        },
        {
          "path_or_url": "output/bekko-renew-audit/retrieval_samples.jsonl",
          "what_verified": "Contains the exact sampled inference inputs, current instructions, candidates, targets, and provenance at the per-sample line numbers listed in sample_refs."
        },
        {
          "path_or_url": "output/synthetic_dataset/nanobeir-en-normalized-v1/scores.jsonl",
          "what_verified": "Original saved teacher records were directly matched for every sampled row (text, source role, raw score, and corrected target): 5/5. Source-positive floor metadata reports 19 corrected document scores in this subset; raw teacher scores remain available."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/judge_prompt_scores_only.txt",
          "what_verified": "The teacher is explicitly asked for each document’s expected marginal usefulness for producing a correct, specific, well-grounded answer if added alone to RAG context; score 3/4 require explicit answer facts."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/score.py",
          "what_verified": "Confirms the source prompt, input construction (query plus documents), score validation, and saved score outputs; source roles and earlier scores are not shown to the judge."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/normalize_scores.py",
          "what_verified": "Keeps original 0–4 teacher values and scales them by 1/4; attaches the answer-evidence rubric."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/fix_positive_scores.py",
          "what_verified": "Raises each source-positive raw 0/1/2 rating to 3, preserves original and corrected scores, and does not alter source-negative scores."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/hakari-bench/NanoBEIR-en",
          "what_verified": "Original source-card reference in the pinned raw-source catalog: NanoSciFact; checked revision not recorded, declared license unknown."
        },
        {
          "path_or_url": "output/bekko-system-one-dataset-v1/dataset_metadata.json",
          "what_verified": "Subset source-task card states category='natural_language'; A compact SciFact split for retrieving scientific abstracts that support or refute claims. It tests evidence retrieval in research literature with scientific terminology. This split is the compact NanoSciFact subset used by NanoBEIR-en.. Current subset is test-only and source positive qrels affect positive-floor correction."
        },
        {
          "path_or_url": "output/bekko-renew-audit/retrieval-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/judge_prompt_scores_only.txt",
          "what_verified": "Original teacher rubric uses integer0–4; native criteria values are0–4. Renew aligns instruction numerals to those unchanged values. The old0–1 presentation remains recoverable through value_scale=0.25 in target metadata; targets and rubric meanings are unchanged."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "0": "Not useful or wrong target.",
            "1": "Topic overlap or weak adjacent context.",
            "2": "Useful context about the exact target without the requested answer.",
            "3": "Explicit facts answering an important part for the exact target.",
            "4": "Direct sufficient answer."
          },
          "id": "relevance",
          "instructions": {
            "mode": "replace",
            "value": "Rate the document independently for its usefulness in answering the query, using the 0–4 criterion levels. Use only the query and document. Choose the lowest fitting level. Levels 3 and 4 require explicit answer facts about the exact target; topical overlap alone is insufficient."
          },
          "selector": {
            "decision_id": "relevance",
            "kind": "judgment",
            "type": "score"
          }
        }
      ],
      "source_profile": "synthetic_relevance__nanobeir__NanoSciFact"
    },
    "synthetic_relevance__nanobeir__NanoTouche2020": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Preserve this exact independent teacher Score task and rubric. The assignment reports 46 source-positive scores raised to >=0.75; where the original teacher gave a lower score, this can overrule the answer-evidence rubric. Treat targets as corrected teacher labels, not human gold or official source-qrel metrics; do not normalize across documents.",
      "evidence": [
        {
          "path_or_url": "output/bekko-system-one-dataset-v1/synthetic_relevance__nanobeir__NanoTouche2020",
          "what_verified": "Loaded this subset from the active release and sampled {'test': 5} rows: index 0, floor(n/2), and n-1 in each split; split populations are {'test': 392}."
        },
        {
          "path_or_url": "output/bekko-renew-audit/retrieval_samples.jsonl",
          "what_verified": "Contains the exact sampled inference inputs, current instructions, candidates, targets, and provenance at the per-sample line numbers listed in sample_refs."
        },
        {
          "path_or_url": "output/synthetic_dataset/nanobeir-en-normalized-v1/scores.jsonl",
          "what_verified": "Original saved teacher records were directly matched for every sampled row (text, source role, raw score, and corrected target): 5/5. Source-positive floor metadata reports 46 corrected document scores in this subset; raw teacher scores remain available."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/judge_prompt_scores_only.txt",
          "what_verified": "The teacher is explicitly asked for each document’s expected marginal usefulness for producing a correct, specific, well-grounded answer if added alone to RAG context; score 3/4 require explicit answer facts."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/score.py",
          "what_verified": "Confirms the source prompt, input construction (query plus documents), score validation, and saved score outputs; source roles and earlier scores are not shown to the judge."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/normalize_scores.py",
          "what_verified": "Keeps original 0–4 teacher values and scales them by 1/4; attaches the answer-evidence rubric."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/fix_positive_scores.py",
          "what_verified": "Raises each source-positive raw 0/1/2 rating to 3, preserves original and corrected scores, and does not alter source-negative scores."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/hakari-bench/NanoBEIR-en",
          "what_verified": "Original source-card reference in the pinned raw-source catalog: NanoTouche2020; checked revision not recorded, declared license unknown."
        },
        {
          "path_or_url": "output/bekko-system-one-dataset-v1/dataset_metadata.json",
          "what_verified": "Subset source-task card states category='natural_language'; A compact Touche 2020 argument-retrieval split. Queries are controversial questions, and relevant documents provide arguments that address the topic. This split is the compact NanoTouche2020 subset used by NanoBEIR-en.. Current subset is test-only and source positive qrels affect positive-floor correction."
        },
        {
          "path_or_url": "output/bekko-renew-audit/retrieval-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/judge_prompt_scores_only.txt",
          "what_verified": "Original teacher rubric uses integer0–4; native criteria values are0–4. Renew aligns instruction numerals to those unchanged values. The old0–1 presentation remains recoverable through value_scale=0.25 in target metadata; targets and rubric meanings are unchanged."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "0": "Not useful or wrong target.",
            "1": "Topic overlap or weak adjacent context.",
            "2": "Useful context about the exact target without the requested answer.",
            "3": "Explicit facts answering an important part for the exact target.",
            "4": "Direct sufficient answer."
          },
          "id": "relevance",
          "instructions": {
            "mode": "replace",
            "value": "Rate the document independently for its usefulness in answering the query, using the 0–4 criterion levels. Use only the query and document. Choose the lowest fitting level. Levels 3 and 4 require explicit answer facts about the exact target; topical overlap alone is insufficient."
          },
          "selector": {
            "decision_id": "relevance",
            "kind": "judgment",
            "type": "score"
          }
        }
      ],
      "source_profile": "synthetic_relevance__nanobeir__NanoTouche2020"
    },
    "synthetic_relevance__nanocoir": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Preserve this exact independent teacher Score task and rubric. The assignment reports 22 source-positive scores raised to >=0.75; where the original teacher gave a lower score, this can overrule the answer-evidence rubric. Treat targets as corrected teacher labels, not human gold or official source-qrel metrics; do not normalize across documents.",
      "evidence": [
        {
          "path_or_url": "output/bekko-system-one-dataset-v1/synthetic_relevance__nanocoir",
          "what_verified": "Loaded this subset from the active release and sampled {'test': 5} rows: index 0, floor(n/2), and n-1 in each split; split populations are {'test': 800}."
        },
        {
          "path_or_url": "output/bekko-renew-audit/retrieval_samples.jsonl",
          "what_verified": "Contains the exact sampled inference inputs, current instructions, candidates, targets, and provenance at the per-sample line numbers listed in sample_refs."
        },
        {
          "path_or_url": "output/synthetic_dataset/nanocoir-normalized-v1/scores.jsonl",
          "what_verified": "Original saved teacher records were directly matched for every sampled row (text, source role, raw score, and corrected target): 5/5. Source-positive floor metadata reports 22 corrected document scores in this subset; raw teacher scores remain available."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/judge_prompt_scores_only.txt",
          "what_verified": "The teacher is explicitly asked for each document’s expected marginal usefulness for producing a correct, specific, well-grounded answer if added alone to RAG context; score 3/4 require explicit answer facts."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/score.py",
          "what_verified": "Confirms the source prompt, input construction (query plus documents), score validation, and saved score outputs; source roles and earlier scores are not shown to the judge."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/normalize_scores.py",
          "what_verified": "Keeps original 0–4 teacher values and scales them by 1/4; attaches the answer-evidence rubric."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/fix_positive_scores.py",
          "what_verified": "Raises each source-positive raw 0/1/2 rating to 3, preserves original and corrected scores, and does not alter source-negative scores."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/hakari-bench/NanoCoIR",
          "what_verified": "Original source-card reference in the pinned raw-source catalog: NanoApps; checked revision not recorded, declared license unknown."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/hakari-bench/NanoCoIR",
          "what_verified": "Original source-card reference in the pinned raw-source catalog: NanoCodeFeedbackMT; checked revision not recorded, declared license unknown."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/hakari-bench/NanoCoIR",
          "what_verified": "Original source-card reference in the pinned raw-source catalog: NanoCodeFeedbackST; checked revision not recorded, declared license unknown."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/hakari-bench/NanoCoIR",
          "what_verified": "Original source-card reference in the pinned raw-source catalog: NanoCodeSearchNet; checked revision not recorded, declared license unknown."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/hakari-bench/NanoCoIR",
          "what_verified": "Original source-card reference in the pinned raw-source catalog: NanoCodeSearchNetCCR; checked revision not recorded, declared license unknown."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/hakari-bench/NanoCoIR",
          "what_verified": "Original source-card reference in the pinned raw-source catalog: NanoCodeTransOceanContest; checked revision not recorded, declared license unknown."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/hakari-bench/NanoCoIR",
          "what_verified": "Original source-card reference in the pinned raw-source catalog: NanoCodeTransOceanDL; checked revision not recorded, declared license unknown."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/hakari-bench/NanoCoIR",
          "what_verified": "Original source-card reference in the pinned raw-source catalog: NanoCosQA; checked revision not recorded, declared license unknown."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/hakari-bench/NanoCoIR",
          "what_verified": "Original source-card reference in the pinned raw-source catalog: NanoStackOverflowQA; checked revision not recorded, declared license unknown."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/hakari-bench/NanoCoIR",
          "what_verified": "Original source-card reference in the pinned raw-source catalog: NanoSyntheticText2SQL; checked revision not recorded, declared license unknown."
        },
        {
          "path_or_url": "output/bekko-system-one-dataset-v1/dataset_metadata.json",
          "what_verified": "Subset source-task card states category=None; . Current subset is test-only and source positive qrels affect positive-floor correction."
        },
        {
          "path_or_url": "output/bekko-renew-audit/retrieval-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        },
        {
          "path_or_url": "scripts/synthetic_dataset/judge_prompt_scores_only.txt",
          "what_verified": "Original teacher rubric uses integer0–4; native criteria values are0–4. Renew aligns instruction numerals to those unchanged values. The old0–1 presentation remains recoverable through value_scale=0.25 in target metadata; targets and rubric meanings are unchanged."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "0": "Not useful or wrong target.",
            "1": "Topic overlap or weak adjacent context.",
            "2": "Useful context about the exact target without the requested answer.",
            "3": "Explicit facts answering an important part for the exact target.",
            "4": "Direct sufficient answer."
          },
          "id": "relevance",
          "instructions": {
            "mode": "replace",
            "value": "Rate the document independently for its usefulness in answering the query, using the 0–4 criterion levels. Use only the query and document. Choose the lowest fitting level. Levels 3 and 4 require explicit answer facts about the exact target; topical overlap alone is insufficient."
          },
          "selector": {
            "decision_id": "relevance",
            "kind": "judgment",
            "type": "score"
          }
        }
      ],
      "source_profile": "synthetic_relevance__nanocoir"
    },
    "temporal_nli": {
      "disposition": "include",
      "disposition_reason": "Reviewed English typed judgment conversion; preserve selected source splits and label semantics.; For entailment-style Noul rows, false means the queried support/entailment relation is not established; do not relabel all false examples as contradictions.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/tasksource/temporal-nli/blob/d5cedbbdb9f1e7591569ebaf7cf1dd238f0b624b/README.md",
          "what_verified": "Upstream dataset/card identifier tasksource/temporal-nli at pinned revision d5cedbbdb9f1e7591569ebaf7cf1dd238f0b624b; license metadata is ['apache-2.0']. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/tasksource/temporal-nli/raw/d5cedbbdb9f1e7591569ebaf7cf1dd238f0b624b/README.md",
          "what_verified": "Primary source card was fetched and read at d5cedbbdb9f1e7591569ebaf7cf1dd238f0b624b; card excerpt: --- license: apache-2.0 --- ``` @inproceedings{thukral-etal-2021-probing, title = \"Probing Language Models for Understanding of Temporal Expressions\", author = \"Thukral, Shivin  and"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/temporal_nli/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 2 source artifacts; 2 referenced local cache paths exist and integrity digests match. Plan source revision d5cedbbdb9f1e7591569ebaf7cf1dd238f0b624b."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/temporal_nli/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/temporal_nli",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 97721 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 6 actual rows from 2 pinned raw source files for source key temporal_nli; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "id": "temporal_nli__choice__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Preserve source Choice option IDs/descriptions and order; sampled option lists can be instance-specific.",
          "selector": {
            "decision_id": "relation",
            "kind": "judgment",
            "type": "choice"
          }
        }
      ],
      "source_profile": "temporal_nli"
    },
    "tracie": {
      "disposition": "include",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching.",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "4967d8aaa06a4ea97a88bf77071d66be0140bc45",
          "sha256": "5ccc003829508d891fc1dfa45661d78d0817eabe828d0dd813d51f60747f189d",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/tasksource/tracie/4967d8aaa06a4ea97a88bf77071d66be0140bc45/README.md?%2Fdatasets%2Ftasksource%2Ftracie%2Fresolve%2F4967d8aaa06a4ea97a88bf77071d66be0140bc45%2FREADME.md=&etag=%224400eb027fe878a2a073ef59bfe88043fc77b380%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "tracie:default:test:0",
          "local_path": "output/typed_decisions/additional20-v1/cache/85231c7c706fa21ea709d7db607caab2082cdf6a7241090e222ab7a5b522a803",
          "original_row_match_status": "matched",
          "sha256": "5519bb6512463baef1cfedd9854417dfb5989c6e08ac6f706cc69425d1b4207f",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://huggingface.co/datasets/tasksource/tracie/resolve/dc7026e1125db54d1d16b7b22495aacb214cc397/default/test/0000.parquet"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/additional_adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "false": "The story does not support the stated temporal relation; lack of support does not establish the opposite relation.",
            "true": "The story supports the stated temporal relation between events."
          },
          "id": "entails",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "",
          "selector": {
            "decision_id": "entails",
            "type": "noul"
          }
        }
      ],
      "source_profile": "tracie"
    },
    "laya__typed_decisions": {
      "disposition": "include",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching.",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "c76749ec58bd8c3d2ea706b31c333a9059c38f90",
          "sha256": "b530a260ffbff42b73205f055bd88ee28300304aaf564bc60ea3e5f52035f5e4",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/LocalLLaMA/typed-decisions/c76749ec58bd8c3d2ea706b31c333a9059c38f90/README.md?%2Fdatasets%2FLocalLLaMA%2Ftyped-decisions%2Fresolve%2Fc76749ec58bd8c3d2ea706b31c333a9059c38f90%2FREADME.md=&etag=%22f2c50bf989cb5678b0665fb50635929f7ac193bd%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "typed_decisions:all:test:0",
          "local_path": "output/typed_decisions/laya-v1/cases/cache/4219cfef302da4986d10cd385f6b51230eeb99c68d4beef849e80c87ed5c97e3",
          "original_row_match_status": "matched",
          "sha256": "4f294f218ea1da27f3efef936359389c62ea4d3973a41457732990f1d31b647c",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://huggingface.co/datasets/LocalLLaMA/typed-decisions/resolve/d2d524a06124415749828ab818dd80a82fd612f8/all/test/0000.parquet"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            },
            {
              "path_or_url": "scripts/typed_decision_dataset/laya_adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {},
          "id": "all_native_task_definitions",
          "instructions": {
            "mode": "preserve"
          },
          "preserve_case_specific_criteria": true,
          "preserve_source_rubrics": true,
          "selector": {},
          "task_meaning": "Each row supplies its own question, option criteria or score rubric for one of four workflows. Keep those original row-level definitions and their aligned state; do not replace them with a subset-level shared definition."
        }
      ],
      "source_profile": "laya__typed_decisions"
    },
    "ud_ewt": {
      "disposition": "include_with_caveat",
      "disposition_reason": "See insights/2026-09-24-additional20-dataset-build.md for source license, construction and typed-label review.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/universal-dependencies/universal_dependencies/blob/c3fed78df6b9f8cd27ed24cc7142e9eccf1e6703/README.md",
          "what_verified": "Upstream dataset/card identifier universal-dependencies/universal_dependencies at pinned revision c3fed78df6b9f8cd27ed24cc7142e9eccf1e6703; license metadata is ['cc-by-sa-4.0']. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/universal-dependencies/universal_dependencies/raw/c3fed78df6b9f8cd27ed24cc7142e9eccf1e6703/README.md",
          "what_verified": "Primary source card was fetched and read at c3fed78df6b9f8cd27ed24cc7142e9eccf1e6703; card excerpt: ## Dataset Card (v2.0) for Universal Dependencies Treebank **Version 2.0.0** introduces significant improvements and breaking changes: - **Parquet Format:** faster loading with HuggingFace datasets >=4.0.0 - **MWT Support:** New `mwt` field provides structured multi-word token information"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/ud_ewt/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 3 source artifacts; 3 referenced local cache paths exist and integrity digests match. Plan source revision c3fed78df6b9f8cd27ed24cc7142e9eccf1e6703."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/ud_ewt/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/ud_ewt",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 15520 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 9 actual rows from 3 pinned raw source files for source key ud_ewt; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "id": "ud_ewt__choice__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Preserve source Choice option IDs/descriptions and order; sampled option lists can be instance-specific.",
          "selector": {
            "decision_id": "[0-9]+",
            "kind": "judgment",
            "type": "choice"
          }
        }
      ],
      "source_profile": "ud_ewt"
    },
    "winobias": {
      "disposition": "include",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching.",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "0bce984dd081bbc10b0622f326727a024c607895",
          "sha256": "db78630bc08e3f38eb60a5ffb12f138e4fef9c9f804a3b18e70eb125dfbf0223",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://raw.githubusercontent.com/uclanlp/corefBias/0bce984dd081bbc10b0622f326727a024c607895/README.md"
        },
        {
          "adapter_replay": "matched",
          "case_id": "winobias:anti_stereotyped_type1.txt.test:test:0",
          "local_path": "output/typed_decisions/additional20-v1/cache/b8f0e7caadfc5b23e142fceacb849907e751a28382ed40c4a116195ee361538f",
          "original_row_match_status": "matched",
          "sha256": "331db5bd74bfefebf146a60b67645152a4a1991570d2a56f57154103ac361dd2",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://raw.githubusercontent.com/uclanlp/corefBias/0bce984dd081bbc10b0622f326727a024c607895/WinoBias/wino/data/anti_stereotyped_type1.txt.test"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/additional_adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {},
          "id": "referent",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "Options and their descriptions vary by case; use the row’s own option-ID/description pairs. Do not assign a fixed meaning to option IDs across rows.",
          "option_semantics": "per_case_candidates",
          "preserve_case_options": true,
          "selector": {
            "decision_id": "referent",
            "type": "choice"
          }
        }
      ],
      "source_profile": "winobias"
    },
    "winowhy": {
      "disposition": "include",
      "disposition_reason": "All available splits were sampled and one pinned raw source row was replayed through the converter with state, decision, option meanings, and target matching.",
      "evidence": [
        {
          "note": "Full fetched card is retained in typed_a-original-source-review.json; a missing pinned card and any unpinned supplemental reference are identified explicitly.",
          "revision": "b7732098d32396fb9f690e9d50cb7f1758861305",
          "sha256": "3d0c031c9d6aeefa4bcdf76ba316ba41b8107c3df25656eb975619cdee95c87d",
          "source": "pinned_original_card",
          "status": "read_pinned",
          "supplemental_original_source_sha256": null,
          "supplemental_original_source_url": null,
          "url": "https://huggingface.co/api/resolve-cache/datasets/tasksource/winowhy/b7732098d32396fb9f690e9d50cb7f1758861305/README.md?%2Fdatasets%2Ftasksource%2Fwinowhy%2Fresolve%2Fb7732098d32396fb9f690e9d50cb7f1758861305%2FREADME.md=&etag=%224c274b04bb572df1ad32e1b15e1f332d56445e2a%22"
        },
        {
          "adapter_replay": "matched",
          "case_id": "winowhy:default:train:0",
          "local_path": "output/typed_decisions/permissive-v1/cache/d72bc8ca712b02ead3f33eae9ecc8a71ac9284429160dad0f242285d72b75664",
          "original_row_match_status": "matched",
          "sha256": "68393d365cbb2b5121c81bd0a823de9778c66cc23aeb3f89d86f78f423ca9739",
          "source": "pinned_raw_example",
          "source_row_index": 0,
          "state_and_decisions_matched": true,
          "url": "https://huggingface.co/datasets/tasksource/winowhy/resolve/f37a583a2172eff7eed294d8d94faf17756824a2/default/train/0000.parquet"
        },
        {
          "paths": [
            {
              "path_or_url": "scripts/typed_decision_dataset/permissive_adapters.py",
              "what_verified": "Read the relevant conversion branch to trace source fields into state, instructions, choices/rubric, decision IDs, and targets."
            }
          ],
          "source": "converter"
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_a-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "criteria_by_id": {
            "false": "The supplied reason does not justify the pronoun interpretation; this does not by itself prove the interpretation wrong.",
            "true": "The supplied reason justifies the pronoun interpretation in the sentence."
          },
          "id": "valid_reason",
          "instructions": {
            "mode": "preserve"
          },
          "notes": "",
          "selector": {
            "decision_id": "valid_reason",
            "type": "noul"
          }
        }
      ],
      "source_profile": "winowhy"
    },
    "wiqa": {
      "disposition": "include",
      "disposition_reason": "Official complete v2 archive identified by pinned HF loader, downloaded without executing code and pinned by SHA-256. Includes process steps missing from tasksource mirror.",
      "evidence": [
        {
          "path_or_url": "https://huggingface.co/datasets/allenai/wiqa/blob/8dda4b5e237452fb939b39326f68ad6607e75ab6/README.md",
          "what_verified": "Upstream dataset/card identifier allenai/wiqa at pinned revision 8dda4b5e237452fb939b39326f68ad6607e75ab6; license metadata is ['apache-2.0']. Local source plan/report provides downloaded-file trace."
        },
        {
          "path_or_url": "https://huggingface.co/datasets/allenai/wiqa/raw/8dda4b5e237452fb939b39326f68ad6607e75ab6/README.md",
          "what_verified": "Primary source card was fetched and read at 8dda4b5e237452fb939b39326f68ad6607e75ab6; card excerpt: # Dataset Card for \"wiqa\" ## Table of Contents - [Dataset Description](#dataset-description) - [Dataset Summary](#dataset-summary)"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/wiqa/upstream-source-source-plan.json",
          "what_verified": "Pinned upstream file inventory / SHA-256 trace; 1 source artifacts; 1 referenced local cache paths exist and integrity digests match. Plan source revision 8dda4b5e237452fb939b39326f68ad6607e75ab6."
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/wiqa/upstream-source-report.json",
          "what_verified": "Local source acquisition/conversion report at source revision None; report status complete"
        },
        {
          "path_or_url": "output/typed_decisions/system-ichi-v1/wiqa",
          "what_verified": "Loaded immutable Arrow export (FollowIR: each followir_* child release); export report has 39087 source cases when reported. Sampled examples span splits and preserve source IDs/instructions/options/targets."
        },
        {
          "path_or_url": "scripts/typed_decision_dataset/adapters.py",
          "what_verified": "Adapter inspected for source-to-task mapping and label conversion."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-raw-samples.json",
          "what_verified": "Inspected 3 actual rows from 1 pinned raw source files for source key wiqa; sample records retain pinned file URL and local cache path."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-native-samples.json",
          "what_verified": "Sampled current immutable native release examples per split/config; verified current task IDs, types, instructions_json/criteria, system prompt, target and case/provenance fields."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-source-integrity.json",
          "what_verified": "Raw source cache SHA-256 integrity audit; source key details recorded in the companion report."
        },
        {
          "path_or_url": "output/bekko-renew-audit/typed_b-review.json",
          "what_verified": "Source and native sample audit; see companion audit artifacts for exact coverage."
        }
      ],
      "rules": [
        {
          "id": "wiqa__choice__000",
          "instructions": {
            "mode": "preserve"
          },
          "review_note": "Preserve source Choice option IDs/descriptions and order; sampled option lists can be instance-specific.",
          "selector": {
            "decision_id": "effect",
            "kind": "judgment",
            "type": "choice"
          }
        }
      ],
      "source_profile": "wiqa"
    },
    "s1mb-generalization-diverse-score": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Author-intended synthetic evaluation, not independently human-validated.",
      "evidence": [
        {
          "path_or_url": "generalization/s1mb-generalization-diverse-score/REVIEW.md",
          "what_verified": "Authored review and source rows retained; preserve task-specific instructions and criteria."
        }
      ],
      "rules": [
        {
          "id": "preserve_authored_generalization",
          "selector": {},
          "instructions": {
            "mode": "preserve"
          }
        }
      ],
      "source_profile": "s1mb-generalization-diverse-score"
    },
    "s1mb-generalization-diverse-noul": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Author-intended synthetic evaluation, not independently human-validated.",
      "evidence": [
        {
          "path_or_url": "generalization/s1mb-generalization-diverse-noul/REVIEW.md",
          "what_verified": "Authored review and source rows retained; preserve task-specific instructions and criteria."
        }
      ],
      "rules": [
        {
          "id": "preserve_authored_generalization",
          "selector": {},
          "instructions": {
            "mode": "preserve"
          }
        }
      ],
      "source_profile": "s1mb-generalization-diverse-noul"
    },
    "s1mb-generalization-diverse-choice": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Author-intended synthetic evaluation, not independently human-validated.",
      "evidence": [
        {
          "path_or_url": "generalization/s1mb-generalization-diverse-choice/REVIEW.md",
          "what_verified": "Authored review and source rows retained; preserve task-specific instructions and criteria."
        }
      ],
      "rules": [
        {
          "id": "preserve_authored_generalization",
          "selector": {},
          "instructions": {
            "mode": "preserve"
          }
        }
      ],
      "source_profile": "s1mb-generalization-diverse-choice"
    },
    "s1mb-generalization-contextual-score": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Author-intended synthetic evaluation, not independently human-validated.",
      "evidence": [
        {
          "path_or_url": "generalization/s1mb-generalization-contextual-score/REVIEW.md",
          "what_verified": "Authored review and source rows retained; preserve task-specific instructions and criteria."
        }
      ],
      "rules": [
        {
          "id": "preserve_authored_generalization",
          "selector": {},
          "instructions": {
            "mode": "preserve"
          }
        }
      ],
      "source_profile": "s1mb-generalization-contextual-score"
    },
    "s1mb-generalization-contextual-noul": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Author-intended synthetic evaluation, not independently human-validated.",
      "evidence": [
        {
          "path_or_url": "generalization/s1mb-generalization-contextual-noul/REVIEW.md",
          "what_verified": "Authored review and source rows retained; preserve task-specific instructions and criteria."
        }
      ],
      "rules": [
        {
          "id": "preserve_authored_generalization",
          "selector": {},
          "instructions": {
            "mode": "preserve"
          }
        }
      ],
      "source_profile": "s1mb-generalization-contextual-noul"
    },
    "s1mb-generalization-contextual-choice": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Author-intended synthetic evaluation, not independently human-validated.",
      "evidence": [
        {
          "path_or_url": "generalization/s1mb-generalization-contextual-choice/REVIEW.md",
          "what_verified": "Authored review and source rows retained; preserve task-specific instructions and criteria."
        }
      ],
      "rules": [
        {
          "id": "preserve_authored_generalization",
          "selector": {},
          "instructions": {
            "mode": "preserve"
          }
        }
      ],
      "source_profile": "s1mb-generalization-contextual-choice"
    },
    "open_jev__amount-extraction-control-v1": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Synthetic reference labels; no general-capability or model-performance claim.",
      "evidence": [
        {
          "url": "https://huggingface.co/datasets/ZefanCai/Open-Jev/tree/c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "revision": "c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "what_verified": "All available source splits, record projection, schema, target alignment, native rendering and persisted readback; semantic labels not exhaustively rederived."
        }
      ],
      "rules": [
        {
          "id": "open_jev_choice",
          "selector": {
            "kind": "judgment",
            "type": "choice"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "per_case_candidates",
          "numeric_scale": null
        },
        {
          "id": "open_jev_noul",
          "selector": {
            "kind": "judgment",
            "type": "noul"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "explicit_no_yes",
          "numeric_scale": null
        }
      ],
      "source_profile": "open_jev__amount-extraction-control-v1"
    },
    "open_jev__browser-control-v1": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Synthetic reference labels; no general-capability or model-performance claim.",
      "evidence": [
        {
          "url": "https://huggingface.co/datasets/ZefanCai/Open-Jev/tree/c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "revision": "c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "what_verified": "All available source splits, record projection, schema, target alignment, native rendering and persisted readback; semantic labels not exhaustively rederived."
        }
      ],
      "rules": [
        {
          "id": "open_jev_choice",
          "selector": {
            "kind": "judgment",
            "type": "choice"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "per_case_candidates",
          "numeric_scale": null
        }
      ],
      "source_profile": "open_jev__browser-control-v1"
    },
    "open_jev__citation-control-v1": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Synthetic reference labels; no general-capability or model-performance claim.",
      "evidence": [
        {
          "url": "https://huggingface.co/datasets/ZefanCai/Open-Jev/tree/c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "revision": "c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "what_verified": "All available source splits, record projection, schema, target alignment, native rendering and persisted readback; semantic labels not exhaustively rederived."
        }
      ],
      "rules": [
        {
          "id": "open_jev_choice",
          "selector": {
            "kind": "judgment",
            "type": "choice"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "per_case_candidates",
          "numeric_scale": null
        }
      ],
      "source_profile": "open_jev__citation-control-v1"
    },
    "open_jev__context-retention-control-v1": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Synthetic reference labels; no general-capability or model-performance claim.",
      "evidence": [
        {
          "url": "https://huggingface.co/datasets/ZefanCai/Open-Jev/tree/c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "revision": "c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "what_verified": "All available source splits, record projection, schema, target alignment, native rendering and persisted readback; semantic labels not exhaustively rederived."
        }
      ],
      "rules": [
        {
          "id": "open_jev_noul",
          "selector": {
            "kind": "judgment",
            "type": "noul"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "explicit_no_yes",
          "numeric_scale": null
        }
      ],
      "source_profile": "open_jev__context-retention-control-v1"
    },
    "open_jev__customer-control-v1": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Synthetic reference labels; no general-capability or model-performance claim. Generated conversations are CC0; upstream TypeSafe documentation question wording has no verified license. Included with explicit caveat, not relicensed as CC0.",
      "evidence": [
        {
          "url": "https://huggingface.co/datasets/ZefanCai/Open-Jev/tree/c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "revision": "c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "what_verified": "All available source splits, record projection, schema, target alignment, native rendering and persisted readback; semantic labels not exhaustively rederived."
        }
      ],
      "rules": [
        {
          "id": "open_jev_choice",
          "selector": {
            "kind": "judgment",
            "type": "choice"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "per_case_candidates",
          "numeric_scale": null
        },
        {
          "id": "open_jev_noul",
          "selector": {
            "kind": "judgment",
            "type": "noul"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "explicit_no_yes",
          "numeric_scale": null
        },
        {
          "id": "open_jev_score",
          "selector": {
            "kind": "judgment",
            "type": "score"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "per_case_candidates",
          "numeric_scale": "upstream_score_values"
        }
      ],
      "source_profile": "open_jev__customer-control-v1"
    },
    "open_jev__drone-control-v1": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Synthetic reference labels; no general-capability or model-performance claim.",
      "evidence": [
        {
          "url": "https://huggingface.co/datasets/ZefanCai/Open-Jev/tree/c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "revision": "c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "what_verified": "All available source splits, record projection, schema, target alignment, native rendering and persisted readback; semantic labels not exhaustively rederived."
        }
      ],
      "rules": [
        {
          "id": "open_jev_choice",
          "selector": {
            "kind": "judgment",
            "type": "choice"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "per_case_candidates",
          "numeric_scale": null
        },
        {
          "id": "open_jev_noul",
          "selector": {
            "kind": "judgment",
            "type": "noul"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "explicit_no_yes",
          "numeric_scale": null
        },
        {
          "id": "open_jev_score",
          "selector": {
            "kind": "judgment",
            "type": "score"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "per_case_candidates",
          "numeric_scale": "upstream_score_values"
        }
      ],
      "source_profile": "open_jev__drone-control-v1"
    },
    "open_jev__email-selection-control-v1": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Synthetic reference labels; no general-capability or model-performance claim.",
      "evidence": [
        {
          "url": "https://huggingface.co/datasets/ZefanCai/Open-Jev/tree/c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "revision": "c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "what_verified": "All available source splits, record projection, schema, target alignment, native rendering and persisted readback; semantic labels not exhaustively rederived."
        }
      ],
      "rules": [
        {
          "id": "open_jev_choice",
          "selector": {
            "kind": "judgment",
            "type": "choice"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "per_case_candidates",
          "numeric_scale": null
        },
        {
          "id": "open_jev_noul",
          "selector": {
            "kind": "judgment",
            "type": "noul"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "explicit_no_yes",
          "numeric_scale": null
        }
      ],
      "source_profile": "open_jev__email-selection-control-v1"
    },
    "open_jev__entity-alignment-control-v1": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Synthetic reference labels; no general-capability or model-performance claim.",
      "evidence": [
        {
          "url": "https://huggingface.co/datasets/ZefanCai/Open-Jev/tree/c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "revision": "c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "what_verified": "All available source splits, record projection, schema, target alignment, native rendering and persisted readback; semantic labels not exhaustively rederived."
        }
      ],
      "rules": [
        {
          "id": "open_jev_noul",
          "selector": {
            "kind": "judgment",
            "type": "noul"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "explicit_no_yes",
          "numeric_scale": null
        },
        {
          "id": "open_jev_score",
          "selector": {
            "kind": "judgment",
            "type": "score"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "per_case_candidates",
          "numeric_scale": "upstream_score_values"
        }
      ],
      "source_profile": "open_jev__entity-alignment-control-v1"
    },
    "open_jev__ir-control-v1": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Synthetic reference labels; no general-capability or model-performance claim. Choice supervises best-passage selection, not a complete ranking.",
      "evidence": [
        {
          "url": "https://huggingface.co/datasets/ZefanCai/Open-Jev/tree/c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "revision": "c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "what_verified": "All available source splits, record projection, schema, target alignment, native rendering and persisted readback; semantic labels not exhaustively rederived."
        }
      ],
      "rules": [
        {
          "id": "open_jev_choice",
          "selector": {
            "kind": "judgment",
            "type": "choice"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "per_case_candidates",
          "numeric_scale": null
        },
        {
          "id": "open_jev_noul",
          "selector": {
            "kind": "judgment",
            "type": "noul"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "explicit_no_yes",
          "numeric_scale": null
        },
        {
          "id": "open_jev_score",
          "selector": {
            "kind": "judgment",
            "type": "score"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "per_case_candidates",
          "numeric_scale": "upstream_score_values"
        }
      ],
      "source_profile": "open_jev__ir-control-v1"
    },
    "open_jev__mailroom-control-v1": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Synthetic reference labels; no general-capability or model-performance claim.",
      "evidence": [
        {
          "url": "https://huggingface.co/datasets/ZefanCai/Open-Jev/tree/c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "revision": "c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "what_verified": "All available source splits, record projection, schema, target alignment, native rendering and persisted readback; semantic labels not exhaustively rederived."
        }
      ],
      "rules": [
        {
          "id": "open_jev_choice",
          "selector": {
            "kind": "judgment",
            "type": "choice"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "per_case_candidates",
          "numeric_scale": null
        },
        {
          "id": "open_jev_noul",
          "selector": {
            "kind": "judgment",
            "type": "noul"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "explicit_no_yes",
          "numeric_scale": null
        }
      ],
      "source_profile": "open_jev__mailroom-control-v1"
    },
    "open_jev__painting-geometry-v1": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Synthetic reference labels; no general-capability or model-performance claim. Score values use the explicit HSL quantities in criterion descriptions, not upstream ordinal indices; probability vectors are unchanged.",
      "evidence": [
        {
          "url": "https://huggingface.co/datasets/ZefanCai/Open-Jev/tree/c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "revision": "c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "what_verified": "All available source splits, record projection, schema, target alignment, native rendering and persisted readback; semantic labels not exhaustively rederived."
        }
      ],
      "rules": [
        {
          "id": "open_jev_choice",
          "selector": {
            "kind": "judgment",
            "type": "choice"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "per_case_candidates",
          "numeric_scale": null
        },
        {
          "id": "open_jev_noul",
          "selector": {
            "kind": "judgment",
            "type": "noul"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "explicit_no_yes",
          "numeric_scale": null
        },
        {
          "id": "open_jev_score",
          "selector": {
            "kind": "judgment",
            "type": "score"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "per_case_candidates",
          "numeric_scale": "physical_hsl"
        }
      ],
      "source_profile": "open_jev__painting-geometry-v1"
    },
    "open_jev__phone-extraction-control-v1": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Synthetic reference labels; no general-capability or model-performance claim.",
      "evidence": [
        {
          "url": "https://huggingface.co/datasets/ZefanCai/Open-Jev/tree/c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "revision": "c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "what_verified": "All available source splits, record projection, schema, target alignment, native rendering and persisted readback; semantic labels not exhaustively rederived."
        }
      ],
      "rules": [
        {
          "id": "open_jev_choice",
          "selector": {
            "kind": "judgment",
            "type": "choice"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "per_case_candidates",
          "numeric_scale": null
        },
        {
          "id": "open_jev_noul",
          "selector": {
            "kind": "judgment",
            "type": "noul"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "explicit_no_yes",
          "numeric_scale": null
        }
      ],
      "source_profile": "open_jev__phone-extraction-control-v1"
    },
    "open_jev__reasoning-control-v1": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Synthetic reference labels; no general-capability or model-performance claim.",
      "evidence": [
        {
          "url": "https://huggingface.co/datasets/ZefanCai/Open-Jev/tree/c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "revision": "c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "what_verified": "All available source splits, record projection, schema, target alignment, native rendering and persisted readback; semantic labels not exhaustively rederived."
        }
      ],
      "rules": [
        {
          "id": "open_jev_choice",
          "selector": {
            "kind": "judgment",
            "type": "choice"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "per_case_candidates",
          "numeric_scale": null
        },
        {
          "id": "open_jev_noul",
          "selector": {
            "kind": "judgment",
            "type": "noul"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "explicit_no_yes",
          "numeric_scale": null
        },
        {
          "id": "open_jev_score",
          "selector": {
            "kind": "judgment",
            "type": "score"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "per_case_candidates",
          "numeric_scale": "upstream_score_values"
        }
      ],
      "source_profile": "open_jev__reasoning-control-v1"
    },
    "open_jev__silent-failure-control-v1": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Synthetic reference labels; no general-capability or model-performance claim.",
      "evidence": [
        {
          "url": "https://huggingface.co/datasets/ZefanCai/Open-Jev/tree/c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "revision": "c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "what_verified": "All available source splits, record projection, schema, target alignment, native rendering and persisted readback; semantic labels not exhaustively rederived."
        }
      ],
      "rules": [
        {
          "id": "open_jev_noul",
          "selector": {
            "kind": "judgment",
            "type": "noul"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "explicit_no_yes",
          "numeric_scale": null
        }
      ],
      "source_profile": "open_jev__silent-failure-control-v1"
    },
    "open_jev__snake-v1": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Synthetic reference labels; no general-capability or model-performance claim.",
      "evidence": [
        {
          "url": "https://huggingface.co/datasets/ZefanCai/Open-Jev/tree/c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "revision": "c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "what_verified": "All available source splits, record projection, schema, target alignment, native rendering and persisted readback; semantic labels not exhaustively rederived."
        }
      ],
      "rules": [
        {
          "id": "open_jev_choice",
          "selector": {
            "kind": "judgment",
            "type": "choice"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "per_case_candidates",
          "numeric_scale": null
        },
        {
          "id": "open_jev_noul",
          "selector": {
            "kind": "judgment",
            "type": "noul"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "explicit_no_yes",
          "numeric_scale": null
        }
      ],
      "source_profile": "open_jev__snake-v1"
    },
    "open_jev__sponsor-segment-control-v1": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Synthetic reference labels; no general-capability or model-performance claim.",
      "evidence": [
        {
          "url": "https://huggingface.co/datasets/ZefanCai/Open-Jev/tree/c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "revision": "c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "what_verified": "All available source splits, record projection, schema, target alignment, native rendering and persisted readback; semantic labels not exhaustively rederived."
        }
      ],
      "rules": [
        {
          "id": "open_jev_choice",
          "selector": {
            "kind": "judgment",
            "type": "choice"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "per_case_candidates",
          "numeric_scale": null
        }
      ],
      "source_profile": "open_jev__sponsor-segment-control-v1"
    },
    "open_jev__tic_tac_toe-v1": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Synthetic reference labels; no general-capability or model-performance claim. Upstream provides no OOD split for this subset.",
      "evidence": [
        {
          "url": "https://huggingface.co/datasets/ZefanCai/Open-Jev/tree/c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "revision": "c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "what_verified": "All available source splits, record projection, schema, target alignment, native rendering and persisted readback; semantic labels not exhaustively rederived."
        }
      ],
      "rules": [
        {
          "id": "open_jev_choice",
          "selector": {
            "kind": "judgment",
            "type": "choice"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "per_case_candidates",
          "numeric_scale": null
        }
      ],
      "source_profile": "open_jev__tic_tac_toe-v1"
    },
    "open_jev__vizdoom-basic-v1": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Synthetic reference labels; no general-capability or model-performance claim.",
      "evidence": [
        {
          "url": "https://huggingface.co/datasets/ZefanCai/Open-Jev/tree/c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "revision": "c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "what_verified": "All available source splits, record projection, schema, target alignment, native rendering and persisted readback; semantic labels not exhaustively rederived."
        }
      ],
      "rules": [
        {
          "id": "open_jev_noul",
          "selector": {
            "kind": "judgment",
            "type": "noul"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "explicit_no_yes",
          "numeric_scale": null
        },
        {
          "id": "open_jev_score",
          "selector": {
            "kind": "judgment",
            "type": "score"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "per_case_candidates",
          "numeric_scale": "upstream_score_values"
        }
      ],
      "source_profile": "open_jev__vizdoom-basic-v1",
      "excluded_tasks": {
        "choice": "The requested momentum-aware script policy uses a velocity multiplier of 8 and a lateral-error deadband of 10 that are not disclosed in the inference input. Labels measure a hidden weak-script convention, not a uniquely specified action."
      }
    },
    "open_jev__workflow-controls-v1__agent_trace_observability": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Synthetic reference labels; no general-capability or model-performance claim.",
      "evidence": [
        {
          "url": "https://huggingface.co/datasets/ZefanCai/Open-Jev/tree/c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "revision": "c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "what_verified": "All available source splits, record projection, schema, target alignment, native rendering and persisted readback; semantic labels not exhaustively rederived."
        }
      ],
      "rules": [
        {
          "id": "open_jev_noul",
          "selector": {
            "kind": "judgment",
            "type": "noul"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "explicit_no_yes",
          "numeric_scale": null
        }
      ],
      "source_profile": "open_jev__workflow-controls-v1__agent_trace_observability"
    },
    "open_jev__workflow-controls-v1__customer_service": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Synthetic reference labels; no general-capability or model-performance claim. Original within-split repeated inputs are retained, not silently deduplicated; no conflicting targets were found in the frozen input audit.",
      "evidence": [
        {
          "url": "https://huggingface.co/datasets/ZefanCai/Open-Jev/tree/c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "revision": "c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "what_verified": "All available source splits, record projection, schema, target alignment, native rendering and persisted readback; semantic labels not exhaustively rederived."
        }
      ],
      "rules": [
        {
          "id": "open_jev_noul",
          "selector": {
            "kind": "judgment",
            "type": "noul"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "explicit_no_yes",
          "numeric_scale": null
        }
      ],
      "source_profile": "open_jev__workflow-controls-v1__customer_service"
    },
    "open_jev__workflow-controls-v1__invoice_processing": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Synthetic reference labels; no general-capability or model-performance claim. Original within-split repeated inputs are retained, not silently deduplicated; no conflicting targets were found in the frozen input audit.",
      "evidence": [
        {
          "url": "https://huggingface.co/datasets/ZefanCai/Open-Jev/tree/c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "revision": "c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "what_verified": "All available source splits, record projection, schema, target alignment, native rendering and persisted readback; semantic labels not exhaustively rederived."
        }
      ],
      "rules": [
        {
          "id": "open_jev_noul",
          "selector": {
            "kind": "judgment",
            "type": "noul"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "explicit_no_yes",
          "numeric_scale": null
        }
      ],
      "source_profile": "open_jev__workflow-controls-v1__invoice_processing"
    },
    "open_jev__workflow-controls-v1__security_incidents": {
      "disposition": "include_with_caveat",
      "disposition_reason": "Synthetic reference labels; no general-capability or model-performance claim.",
      "evidence": [
        {
          "url": "https://huggingface.co/datasets/ZefanCai/Open-Jev/tree/c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "revision": "c67699e13d0ae25e35b77165a4b6b079bedc8aba",
          "what_verified": "All available source splits, record projection, schema, target alignment, native rendering and persisted readback; semantic labels not exhaustively rederived."
        }
      ],
      "rules": [
        {
          "id": "open_jev_noul",
          "selector": {
            "kind": "judgment",
            "type": "noul"
          },
          "instructions": {
            "mode": "preserve"
          },
          "option_semantics": "explicit_no_yes",
          "numeric_scale": null
        }
      ],
      "source_profile": "open_jev__workflow-controls-v1__security_incidents"
    }
  },
  "relevance_dual_version": "bekko.relevance_dual.v1.20260928"
}