Publish verified typed-decision dataset with source attribution

README.md

@@ -455,21 +455,11 @@ configs:
   - split: test
     path:
     - data/synthetic_relevance__nanobeir__NanoArguAna/test/00000-of-00001.parquet
-- config_name: synthetic_relevance__nanobeir__NanoClimateFEVER
-  data_files:
-  - split: test
-    path:
-    - data/synthetic_relevance__nanobeir__NanoClimateFEVER/test/00000-of-00001.parquet
 - config_name: synthetic_relevance__nanobeir__NanoDBPedia
   data_files:
   - split: test
     path:
     - data/synthetic_relevance__nanobeir__NanoDBPedia/test/00000-of-00001.parquet
-- config_name: synthetic_relevance__nanobeir__NanoFEVER
-  data_files:
-  - split: test
-    path:
-    - data/synthetic_relevance__nanobeir__NanoFEVER/test/00000-of-00001.parquet
 - config_name: synthetic_relevance__nanobeir__NanoFiQA2018
   data_files:
   - split: test
@@ -495,21 +485,11 @@ configs:
   - split: test
     path:
     - data/synthetic_relevance__nanobeir__NanoNQ/test/00000-of-00001.parquet
-- config_name: synthetic_relevance__nanobeir__NanoQuoraRetrieval
-  data_files:
-  - split: test
-    path:
-    - data/synthetic_relevance__nanobeir__NanoQuoraRetrieval/test/00000-of-00001.parquet
 - config_name: synthetic_relevance__nanobeir__NanoSCIDOCS
   data_files:
   - split: test
     path:
     - data/synthetic_relevance__nanobeir__NanoSCIDOCS/test/00000-of-00001.parquet
-- config_name: synthetic_relevance__nanobeir__NanoSciFact
-  data_files:
-  - split: test
-    path:
-    - data/synthetic_relevance__nanobeir__NanoSciFact/test/00000-of-00001.parquet
 - config_name: synthetic_relevance__nanobeir__NanoTouche2020
   data_files:
   - split: test
@@ -572,7 +552,7 @@ dataset = load_dataset("hotchpotch/s1mb-dataset", "laya__typed_decisions", token
 test = dataset["test"]
 ```
 
-Select a configuration explicitly. Parquet uses Zstandard compression and dictionary encoding; all decoded values and row order match the local Arrow release. Only `training-manifest.json` defines active membership. Configurations listed in `quarantine-manifest.json` are retained for audit and must not enter evaluation. Source attribution and license scope are retained in `sources.json`, `SOURCES.md`, `dataset_metadata.json` and `_sources/`. README has one leading YAML block listing every config and split. Hub manifests point to Parquet files; `_sources/local-arrow/` preserves the historical Arrow manifests.
+Select a configuration explicitly. Parquet uses Zstandard compression and dictionary encoding; all decoded values and row order match the local Arrow release. Only `training-manifest.json` defines active membership. Parquet files listed in `quarantine-manifest.json` are retained for audit, excluded from named configurations, and must not enter evaluation. Source attribution and license scope are retained in `sources.json`, `SOURCES.md`, `dataset_metadata.json` and `_sources/`. README has one leading YAML block listing every config and split. Hub manifests point to Parquet files; `_sources/local-arrow/` preserves the historical Arrow manifests.
 
 
 Typed decisions over shared state, with Choice, Noul and numeric Score supervision.

_build/publish_hf.py

@@ -48,6 +48,8 @@ def prepare(source, output, repo_id):
     source_checksums = sha(source / "checksums.json")
     manifest = json.loads((source / "all-data-manifest.json").read_text())
     names = {e["dataset"] for entries in manifest.values() for e in entries}
+    active_manifest = json.loads((source / "training-manifest.json").read_text())
+    active_paths = {e["path"] for entries in active_manifest.values() for e in entries}
     output.mkdir(parents=True)
     # Carry source attribution, license notices and reproducibility records, not duplicate Arrow data.
     for path in source.iterdir():
@@ -105,7 +107,8 @@ def prepare(source, output, repo_id):
                 entry, path=str(directory.relative_to(output)), format="parquet", data_files=files
             )
             published[role].append(item)
-            configs[entry["dataset"]].append({"split": entry["split"], "path": files})
+            if entry["path"] in active_paths:
+                configs[entry["dataset"]].append({"split": entry["split"], "path": files})
             receipts[entry["path"]] = {
                 "cases": len(dataset),
                 "decisions": entry["decisions"],
@@ -173,7 +176,8 @@ def prepare(source, output, repo_id):
         "compression": "zstd",
         "source_checksums_sha256": source_checksums,
         "configs": len(configs),
-        "splits": sum(map(len, published.values())),
+        "splits": sum(map(len, configs.values())),
+        "stored_splits": sum(map(len, published.values())),
         "parquet_bytes": sum(r["bytes"] for r in receipts.values()),
         "receipts": receipts,
     }
@@ -210,8 +214,8 @@ def prepare(source, output, repo_id):
         'test = dataset["test"]\n```\n\n'
         "Select a configuration explicitly. Parquet uses Zstandard compression and dictionary encoding; "
         "all decoded values and row order match the local Arrow release. "
-        "Only `training-manifest.json` defines active membership. Configurations listed in "
-        "`quarantine-manifest.json` are retained for audit and must not enter evaluation. "
+        "Only `training-manifest.json` defines active membership. Parquet files listed in "
+        "`quarantine-manifest.json` are retained for audit, excluded from named configurations, and must not enter evaluation. "
         "Source attribution and license scope are retained in `sources.json`, `SOURCES.md`, "
         "`dataset_metadata.json` and `_sources/`. README has one leading YAML block listing every config and split. "
         "Hub manifests point to Parquet files; `_sources/local-arrow/` preserves the historical Arrow manifests.\n\n"

_source/_additions/synthetic_gpt6astra/synthetic_gpt6astra_choice/dataset_metadata.json

@@ -1,29 +1,3 @@
-{
-  "name": "S1MB-choice-synthetic-dataset-gtp6astra",
-  "task": "choice",
-  "schema_version": "system_one.v1",
-  "version": "english-v1",
-  "created_date": "2026-09-27",
-  "author": "GPT-6 Astra / Codex",
-  "evaluation_only": true,
-  "language": "en",
-  "label_type": "author-intended one-hot",
-  "datasets": {
-    "authored_choice_100": {
-      "task_types": [
-        "choice"
-      ],
-      "splits": {
-        "test": 100
-      }
-    }
-  },
-  "source_sha256": {
-    "experiments/synthetic_typed_gpt6astra/english_score.py": "af028395417d122c0ead8975943a0dac0b10cfaf5630f439bcf9204e9df8df67",
-    "experiments/synthetic_typed_gpt6astra/choice.py": "4bff0ed5200da11993ddd5b93e69db2150003ddaf1f444ef9d9d584dd7a02adb",
-    "experiments/synthetic_typed_gpt6astra/noul.py": "9a2478e80d3bf46fd5331e70b3eb9a461de7e0b341fb442c616459d98867b0c6",
-    "experiments/synthetic_typed_gpt6astra/build.py": "d0f00752742ed0e657f9846ae751569547aec71848197100674bb80e70cf7f94"
-  },
-  "directory_spelling": "noel and gtp6astra preserve the user-requested naming; schema uses noul",
-  "overlap_policy": "Related and cross-task contexts are intentional; use group_id to keep exact shared contexts together."
-}
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9f012438709526d23cfb1a598b7c4f4aa7a6d449e1e4c0f1b3be92913b72e68
+size 1192

_source/_additions/synthetic_gpt6astra/synthetic_gpt6astra_noul/dataset_metadata.json

@@ -1,29 +1,3 @@
-{
-  "name": "S1MB-Noul-synthetic-dataset-gtp6astra",
-  "task": "noul",
-  "schema_version": "system_one.v1",
-  "version": "english-v1",
-  "created_date": "2026-09-27",
-  "author": "GPT-6 Astra / Codex",
-  "evaluation_only": true,
-  "language": "en",
-  "label_type": "author-intended one-hot",
-  "datasets": {
-    "authored_noul_100": {
-      "task_types": [
-        "noul"
-      ],
-      "splits": {
-        "test": 100
-      }
-    }
-  },
-  "source_sha256": {
-    "experiments/synthetic_typed_gpt6astra/english_score.py": "af028395417d122c0ead8975943a0dac0b10cfaf5630f439bcf9204e9df8df67",
-    "experiments/synthetic_typed_gpt6astra/choice.py": "4bff0ed5200da11993ddd5b93e69db2150003ddaf1f444ef9d9d584dd7a02adb",
-    "experiments/synthetic_typed_gpt6astra/noul.py": "9a2478e80d3bf46fd5331e70b3eb9a461de7e0b341fb442c616459d98867b0c6",
-    "experiments/synthetic_typed_gpt6astra/build.py": "0eac4ace1284b9a04cfb72a5677df40051eb46435c24a17d09df5ee3405e48f1"
-  },
-  "directory_spelling": "Directory uses Noul; the canonical schema type is noul",
-  "overlap_policy": "Related and cross-task contexts are intentional; use group_id to keep exact shared contexts together."
-}
+version https://git-lfs.github.com/spec/v1
+oid sha256:0961671806b0031f68847bc25fef822d70670f90e99b2509331c3e710c34b524
+size 1167

_source/_additions/synthetic_gpt6astra/synthetic_gpt6astra_score/dataset_metadata.json

@@ -1,30 +1,3 @@
-{
-  "name": "S1MB-score-synthetic-dataset-gtp6astra",
-  "task": "score",
-  "schema_version": "system_one.v1",
-  "version": "english-v2",
-  "created_date": "2026-09-27",
-  "author": "GPT-6 Astra / Codex",
-  "evaluation_only": true,
-  "language": "en",
-  "label_type": "author-intended one-hot",
-  "datasets": {
-    "authored_score_100": {
-      "task_types": [
-        "score"
-      ],
-      "splits": {
-        "test": 100
-      }
-    }
-  },
-  "source_sha256": {
-    "experiments/synthetic_typed_gpt6astra/english_score.py": "af028395417d122c0ead8975943a0dac0b10cfaf5630f439bcf9204e9df8df67",
-    "experiments/synthetic_typed_gpt6astra/choice.py": "4bff0ed5200da11993ddd5b93e69db2150003ddaf1f444ef9d9d584dd7a02adb",
-    "experiments/synthetic_typed_gpt6astra/noul.py": "9a2478e80d3bf46fd5331e70b3eb9a461de7e0b341fb442c616459d98867b0c6",
-    "experiments/synthetic_typed_gpt6astra/build.py": "d0f00752742ed0e657f9846ae751569547aec71848197100674bb80e70cf7f94",
-    "experiments/synthetic_score_gpt6astra/author.py": "d65f3310b38e1f45504ac70709e711a7d9d8544808363d7b39538b45b37e08ed"
-  },
-  "directory_spelling": "noel and gtp6astra preserve the user-requested naming; schema uses noul",
-  "overlap_policy": "Related and cross-task contexts are intentional; use group_id to keep exact shared contexts together."
-}
+version https://git-lfs.github.com/spec/v1
+oid sha256:f78bebbcfcd9f5cd4d35eb9ed27e497a04dcd5495bd0faf9b0dcaee85213853c
+size 1311