Publish Ropedia Xperience-10M derived artifacts

PROJECT_README.md

@@ -92,6 +92,7 @@ multi-episode held-out model metrics:
 | Research takeaways | `RESEARCH_TAKEAWAYS.md`, `docs/data/research_takeaways.json`, `scripts/build_research_takeaways.py` | summarizes result interpretation from committed metrics and identifies which experiments need held-out episodes |
 | Research roadmap | `RESEARCH_ROADMAP.md`, `docs/data/research_roadmap.json` | stages the path from public-sample task development to multi-episode held-out evaluation and larger omni-model extensions |
 | 12-task suite | `scripts/episode_task_suite.py`, per-task `metrics.json`, predictions | chronological single-episode split |
+| Single-episode diagnostics | `scripts/single_episode_diagnostics.py`, `results/single_episode_diagnostics/`, `docs/single_episode_explorer.html` | modality ablations, timeline overlay, object-label export, alignment stress tests, and interactive window inspection from one sample episode |
 | Neural heads | `scripts/neural_task_models.py`, `results/episode_task_suite/neural_mlp/` | compact MLP heads, not a foundation model |
 | Research directions | `research_direction_taxonomy.json`, extension probe results | direct/proxy/diagnostic evidence, not full solutions |
 | Task surface integrity | `docs/data/task_surface_integrity.json`, `scripts/validate_task_surface.py` | public task cards stay human-readable, thumbnail-backed, and wired to the scrub/play walkthrough storyboard |
@@ -803,6 +804,27 @@ The strongest single-episode self-supervised signal is cross-modal retrieval:
 motion/IMU/camera features retrieve matching depth/video windows substantially
 better than random.
 
+## Single-Episode Diagnostics and Explorer
+
+While waiting for broader Xperience-10M access, the repo now includes an
+artifact-driven diagnostics pass over the public sample episode:
+
+- `results/single_episode_diagnostics/object_labels/window_object_labels.csv`
+  exports 1,161 real window-level object-label sets from `annotation.hdf5`.
+- `results/single_episode_diagnostics/modality_ablation/ablation_metrics.csv`
+  recomputes all 96 task/modality cells, including object relevance.
+- `results/single_episode_diagnostics/timeline_overlay/timeline_overlay.csv`
+  aligns 2,079 existing prediction rows back to the episode timeline.
+- `results/single_episode_diagnostics/alignment_stress/alignment_shift_metrics.csv`
+  evaluates cross-modal retrieval under explicit time shifts.
+- `docs/single_episode_explorer.html` is a static interactive page for
+  inspecting window labels, objects, predictions, feature-block statistics, and
+  diagnostic scores.
+
+These are single-episode research diagnostics. They are useful for auditing
+task definitions, feature behavior, and model errors before scaling to more
+episodes; they are not reported as multi-episode benchmark results.
+
 ## Reproducibility Check
 
 I re-ran the full pipeline from the local raw public sample into an ignored

README.md

@@ -112,6 +112,11 @@ remain setup-stage evidence; 32-episode held-out metrics require the full pilot.
 
 This is the explorable artifact half of the project. You can inspect the task outputs, compare the committed metrics, and understand the single-episode limitations without downloading the raw videos first.
 
+The latest bundle also includes `docs/single_episode_explorer.html` and
+`docs/data/single_episode_explorer.json`, a static window-level explorer for
+labels, object sets, predictions, feature-block statistics, modality ablations,
+and alignment stress diagnostics.
+
 ## 90-Second Research Project Path
 
 | Step | Question | Primary artifacts |
@@ -124,7 +129,8 @@ This is the explorable artifact half of the project. You can inspect the task ou
 | 6 | How do I reproduce it? | `REPRODUCIBILITY.md`, `docs/data/reproducibility_matrix.json`, `notes/reproducibility_audit.md` |
 | 7 | What is one model input? | `results/episode_task_suite/windows.csv`, `results/episode_task_suite/feature_manifest.json`, `results/episode_task_suite/available_modalities.json` |
 | 8 | Are the task results backed by files? | `results/episode_task_suite/summary_report.json`, `results/episode_task_suite/neural_mlp/`, `docs/data/summary_metrics.json` |
-| 9 | What is still pending? | `results/omni_finetune/DATA_ACCESS_STATUS.md`, `results/omni_finetune/MULTI_EPISODE_ACCESS_STATUS.md`, `scripts/omni/discover_xperience10m_sources.py` |
+| 9 | Can I inspect one window interactively? | `docs/single_episode_explorer.html`, `docs/data/single_episode_explorer.json`, `results/single_episode_diagnostics/` |
+| 10 | What is still pending? | `results/omni_finetune/DATA_ACCESS_STATUS.md`, `results/omni_finetune/MULTI_EPISODE_ACCESS_STATUS.md`, `scripts/omni/discover_xperience10m_sources.py` |
 
 Human-readable artifact guide: `ARTIFACT_GUIDE.md`.
 Project brief: `PROJECT_BRIEF.md` and `docs/data/project_brief.json`.
@@ -202,6 +208,8 @@ Source-of-truth brand asset index: `docs/data/brand_assets.json`.
 - `docs/data/website_integrity.json`: machine-readable website local-reference report
 - `QUALITY_GATES.md` and `docs/data/quality_gates.json`: human-readable and machine-readable release checks
 - `docs/data/live_publication_status.json`: last live public URL verification after upload
+- `docs/single_episode_explorer.html` and `docs/data/single_episode_explorer.json`: static interactive explorer for one sample episode
+- `results/single_episode_diagnostics/`: modality ablations, object-label export, timeline overlay, and alignment stress results
 - `docs/data/project_manifest.json`: machine-readable public URL and citation metadata
 - `docs/data/project_packet.json`: machine-readable project path and scope summary
 - `results/omni_finetune/DATA_ACCESS_STATUS.md`: reader-facing multi-episode data requirement for the Qwen3-Omni pilot

docs/data/mirror_parity.json

@@ -1,9 +1,9 @@
 {
   "status": "pass",
-  "generated_at_utc": "2026-06-03T01:40:55+00:00",
+  "generated_at_utc": "2026-06-03T04:07:33+00:00",
   "hf_root": "hf_publish",
   "summary": {
-    "group_count": 75,
+    "group_count": 88,
     "failure_count": 0,
     "failures_by_surface": {}
   },
@@ -24,6 +24,10 @@
       "name": "repo_hf_website_html_parity",
       "status": "pass"
     },
+    {
+      "name": "repo_hf_diagnostic_result_parity",
+      "status": "pass"
+    },
     {
       "name": "repo_hf_quality_doc_parity",
       "status": "pass"
@@ -377,27 +381,27 @@
       "local": {
         "path": "repo:docs/data/publication_audit.json",
         "exists": true,
-        "bytes": 7096,
-        "sha256": "0c692c42c41414c716d642aa051842d8250b44e57df263d3a962cb5b078f7d09"
+        "bytes": 7097,
+        "sha256": "9ab8df8c416ccdf385c076caee6ad13b3b227e29d9467ee00c1e749c36e0909b"
       },
       "mirrors": {
         "hf_space": {
           "path": "hf_space:data/publication_audit.json",
           "exists": true,
-          "bytes": 7096,
-          "sha256": "0c692c42c41414c716d642aa051842d8250b44e57df263d3a962cb5b078f7d09"
+          "bytes": 7097,
+          "sha256": "9ab8df8c416ccdf385c076caee6ad13b3b227e29d9467ee00c1e749c36e0909b"
         },
         "hf_artifacts": {
           "path": "hf_artifacts:docs/data/publication_audit.json",
           "exists": true,
-          "bytes": 7096,
-          "sha256": "0c692c42c41414c716d642aa051842d8250b44e57df263d3a962cb5b078f7d09"
+          "bytes": 7097,
+          "sha256": "9ab8df8c416ccdf385c076caee6ad13b3b227e29d9467ee00c1e749c36e0909b"
         },
         "hf_model": {
           "path": "hf_model:metrics/publication_audit.json",
           "exists": true,
-          "bytes": 7096,
-          "sha256": "0c692c42c41414c716d642aa051842d8250b44e57df263d3a962cb5b078f7d09"
+          "bytes": 7097,
+          "sha256": "9ab8df8c416ccdf385c076caee6ad13b3b227e29d9467ee00c1e749c36e0909b"
         }
       },
       "failures": []
@@ -681,6 +685,37 @@
       },
       "failures": []
     },
+    {
+      "name": "data/single_episode_explorer.json",
+      "status": "pass",
+      "local": {
+        "path": "repo:docs/data/single_episode_explorer.json",
+        "exists": true,
+        "bytes": 4101241,
+        "sha256": "4ea0e34660de421a1a5ab0a4afcc44edec0ad0fc4e739f13d3af7974163f1897"
+      },
+      "mirrors": {
+        "hf_space": {
+          "path": "hf_space:data/single_episode_explorer.json",
+          "exists": true,
+          "bytes": 4101241,
+          "sha256": "4ea0e34660de421a1a5ab0a4afcc44edec0ad0fc4e739f13d3af7974163f1897"
+        },
+        "hf_artifacts": {
+          "path": "hf_artifacts:docs/data/single_episode_explorer.json",
+          "exists": true,
+          "bytes": 4101241,
+          "sha256": "4ea0e34660de421a1a5ab0a4afcc44edec0ad0fc4e739f13d3af7974163f1897"
+        },
+        "hf_model": {
+          "path": "hf_model:metrics/single_episode_explorer.json",
+          "exists": true,
+          "bytes": 4101241,
+          "sha256": "4ea0e34660de421a1a5ab0a4afcc44edec0ad0fc4e739f13d3af7974163f1897"
+        }
+      },
+      "failures": []
+    },
     {
       "name": "data/source_alignment_audit.json",
       "status": "pass",
@@ -811,27 +846,27 @@
       "local": {
         "path": "repo:docs/data/website_integrity.json",
         "exists": true,
-        "bytes": 12715,
-        "sha256": "b68904e5051827d97f4b28b23fe57e014773c23b1b648a6444391658eed395bf"
+        "bytes": 13149,
+        "sha256": "d813e2b9dab44f290deed5dfd17af9949b6e064723f138117678e6f54e346df7"
       },
       "mirrors": {
         "hf_space": {
           "path": "hf_space:data/website_integrity.json",
           "exists": true,
-          "bytes": 12715,
-          "sha256": "b68904e5051827d97f4b28b23fe57e014773c23b1b648a6444391658eed395bf"
+          "bytes": 13149,
+          "sha256": "d813e2b9dab44f290deed5dfd17af9949b6e064723f138117678e6f54e346df7"
         },
         "hf_artifacts": {
           "path": "hf_artifacts:docs/data/website_integrity.json",
           "exists": true,
-          "bytes": 12715,
-          "sha256": "b68904e5051827d97f4b28b23fe57e014773c23b1b648a6444391658eed395bf"
+          "bytes": 13149,
+          "sha256": "d813e2b9dab44f290deed5dfd17af9949b6e064723f138117678e6f54e346df7"
         },
         "hf_model": {
           "path": "hf_model:metrics/website_integrity.json",
           "exists": true,
-          "bytes": 12715,
-          "sha256": "b68904e5051827d97f4b28b23fe57e014773c23b1b648a6444391658eed395bf"
+          "bytes": 13149,
+          "sha256": "d813e2b9dab44f290deed5dfd17af9949b6e064723f138117678e6f54e346df7"
         }
       },
       "failures": []
@@ -1671,6 +1706,31 @@
       },
       "failures": []
     },
+    {
+      "name": "scripts/build_single_episode_explorer.py",
+      "status": "pass",
+      "local": {
+        "path": "repo:scripts/build_single_episode_explorer.py",
+        "exists": true,
+        "bytes": 29394,
+        "sha256": "c837f4f4a0d7baff8d4e3bea36ea0f669c6a3eb073a0504c0a31bab481a38b73"
+      },
+      "mirrors": {
+        "hf_artifacts": {
+          "path": "hf_artifacts:scripts/build_single_episode_explorer.py",
+          "exists": true,
+          "bytes": 29394,
+          "sha256": "c837f4f4a0d7baff8d4e3bea36ea0f669c6a3eb073a0504c0a31bab481a38b73"
+        },
+        "hf_model": {
+          "path": "hf_model:scripts/build_single_episode_explorer.py",
+          "exists": true,
+          "bytes": 29394,
+          "sha256": "c837f4f4a0d7baff8d4e3bea36ea0f669c6a3eb073a0504c0a31bab481a38b73"
+        }
+      },
+      "failures": []
+    },
     {
       "name": "scripts/build_research_takeaways.py",
       "status": "pass",
@@ -1696,6 +1756,31 @@
       },
       "failures": []
     },
+    {
+      "name": "scripts/single_episode_diagnostics.py",
+      "status": "pass",
+      "local": {
+        "path": "repo:scripts/single_episode_diagnostics.py",
+        "exists": true,
+        "bytes": 57667,
+        "sha256": "865ab2cd732e561fdf006516d50b337878592ab06708f231e9a864f82b3c867f"
+      },
+      "mirrors": {
+        "hf_artifacts": {
+          "path": "hf_artifacts:scripts/single_episode_diagnostics.py",
+          "exists": true,
+          "bytes": 57667,
+          "sha256": "865ab2cd732e561fdf006516d50b337878592ab06708f231e9a864f82b3c867f"
+        },
+        "hf_model": {
+          "path": "hf_model:scripts/single_episode_diagnostics.py",
+          "exists": true,
+          "bytes": 57667,
+          "sha256": "865ab2cd732e561fdf006516d50b337878592ab06708f231e9a864f82b3c867f"
+        }
+      },
+      "failures": []
+    },
     {
       "name": "scripts/verify_live_publication.py",
       "status": "pass",
@@ -1727,21 +1812,21 @@
       "local": {
         "path": "repo:scripts/validate_mirror_parity.py",
         "exists": true,
-        "bytes": 10446,
-        "sha256": "77d1297b7faa1605e50b31e11f30e4f4b8af4b3d566d80bc93d9886295efd15e"
+        "bytes": 11942,
+        "sha256": "730159d3136e1f7fd7db236157d4081096bde0fde5110faeb265eb2724509945"
       },
       "mirrors": {
         "hf_artifacts": {
           "path": "hf_artifacts:scripts/validate_mirror_parity.py",
           "exists": true,
-          "bytes": 10446,
-          "sha256": "77d1297b7faa1605e50b31e11f30e4f4b8af4b3d566d80bc93d9886295efd15e"
+          "bytes": 11942,
+          "sha256": "730159d3136e1f7fd7db236157d4081096bde0fde5110faeb265eb2724509945"
         },
         "hf_model": {
           "path": "hf_model:scripts/validate_mirror_parity.py",
           "exists": true,
-          "bytes": 10446,
-          "sha256": "77d1297b7faa1605e50b31e11f30e4f4b8af4b3d566d80bc93d9886295efd15e"
+          "bytes": 11942,
+          "sha256": "730159d3136e1f7fd7db236157d4081096bde0fde5110faeb265eb2724509945"
         }
       },
       "failures": []
@@ -1952,21 +2037,46 @@
       "local": {
         "path": "repo:docs/index.html",
         "exists": true,
-        "bytes": 159244,
-        "sha256": "5c11e18e9974987d332fc0bdb2ca1dfadfaac98f37362710d1043e58b9074f73"
+        "bytes": 159797,
+        "sha256": "6071494e626ef88821cce4f6f9ee27b7d56011de345add95e1708c921ccc84e4"
       },
       "mirrors": {
         "hf_space": {
           "path": "hf_space:index.html",
           "exists": true,
-          "bytes": 159244,
-          "sha256": "5c11e18e9974987d332fc0bdb2ca1dfadfaac98f37362710d1043e58b9074f73"
+          "bytes": 159797,
+          "sha256": "6071494e626ef88821cce4f6f9ee27b7d56011de345add95e1708c921ccc84e4"
         },
         "hf_artifacts_docs": {
           "path": "hf_artifacts:docs/index.html",
           "exists": true,
-          "bytes": 159244,
-          "sha256": "5c11e18e9974987d332fc0bdb2ca1dfadfaac98f37362710d1043e58b9074f73"
+          "bytes": 159797,
+          "sha256": "6071494e626ef88821cce4f6f9ee27b7d56011de345add95e1708c921ccc84e4"
+        }
+      },
+      "failures": []
+    },
+    {
+      "name": "website/single_episode_explorer.html",
+      "status": "pass",
+      "local": {
+        "path": "repo:docs/single_episode_explorer.html",
+        "exists": true,
+        "bytes": 2641502,
+        "sha256": "e97fede9233ce1329ea113dbbd06d0b6b9e5da986d35a9d6a3e20b626d741935"
+      },
+      "mirrors": {
+        "hf_space": {
+          "path": "hf_space:single_episode_explorer.html",
+          "exists": true,
+          "bytes": 2641502,
+          "sha256": "e97fede9233ce1329ea113dbbd06d0b6b9e5da986d35a9d6a3e20b626d741935"
+        },
+        "hf_artifacts_docs": {
+          "path": "hf_artifacts:docs/single_episode_explorer.html",
+          "exists": true,
+          "bytes": 2641502,
+          "sha256": "e97fede9233ce1329ea113dbbd06d0b6b9e5da986d35a9d6a3e20b626d741935"
         }
       },
       "failures": []
@@ -1996,6 +2106,285 @@
       },
       "failures": []
     },
+    {
+      "name": "results/single_episode_diagnostics/provenance.json",
+      "status": "pass",
+      "local": {
+        "path": "repo:results/single_episode_diagnostics/provenance.json",
+        "exists": true,
+        "bytes": 3962,
+        "sha256": "48087224240941fd92444d4fee94a6146c96c35b5acc429a209db3c5cdb8d24b"
+      },
+      "mirrors": {
+        "hf_space": {
+          "path": "hf_space:results/single_episode_diagnostics/provenance.json",
+          "exists": true,
+          "bytes": 3962,
+          "sha256": "48087224240941fd92444d4fee94a6146c96c35b5acc429a209db3c5cdb8d24b"
+        },
+        "hf_artifacts": {
+          "path": "hf_artifacts:results/single_episode_diagnostics/provenance.json",
+          "exists": true,
+          "bytes": 3962,
+          "sha256": "48087224240941fd92444d4fee94a6146c96c35b5acc429a209db3c5cdb8d24b"
+        },
+        "hf_model": {
+          "path": "hf_model:results/single_episode_diagnostics/provenance.json",
+          "exists": true,
+          "bytes": 3962,
+          "sha256": "48087224240941fd92444d4fee94a6146c96c35b5acc429a209db3c5cdb8d24b"
+        }
+      },
+      "failures": []
+    },
+    {
+      "name": "results/single_episode_diagnostics/README.md",
+      "status": "pass",
+      "local": {
+        "path": "repo:results/single_episode_diagnostics/README.md",
+        "exists": true,
+        "bytes": 1058,
+        "sha256": "e64b13f227f270dc545317636f432fb766d6f7ab8695f69234b2c24b6413a42e"
+      },
+      "mirrors": {
+        "hf_space": {
+          "path": "hf_space:results/single_episode_diagnostics/README.md",
+          "exists": true,
+          "bytes": 1058,
+          "sha256": "e64b13f227f270dc545317636f432fb766d6f7ab8695f69234b2c24b6413a42e"
+        },
+        "hf_artifacts": {
+          "path": "hf_artifacts:results/single_episode_diagnostics/README.md",
+          "exists": true,
+          "bytes": 1058,
+          "sha256": "e64b13f227f270dc545317636f432fb766d6f7ab8695f69234b2c24b6413a42e"
+        },
+        "hf_model": {
+          "path": "hf_model:results/single_episode_diagnostics/README.md",
+          "exists": true,
+          "bytes": 1058,
+          "sha256": "e64b13f227f270dc545317636f432fb766d6f7ab8695f69234b2c24b6413a42e"
+        }
+      },
+      "failures": []
+    },
+    {
+      "name": "results/single_episode_diagnostics/modality_ablation/ablation_metrics.csv",
+      "status": "pass",
+      "local": {
+        "path": "repo:results/single_episode_diagnostics/modality_ablation/ablation_metrics.csv",
+        "exists": true,
+        "bytes": 24871,
+        "sha256": "8e43bfa9b90544d7b09e207de26d1b852dd115b3e7c957eaa4b45e520dc050b8"
+      },
+      "mirrors": {
+        "hf_space": {
+          "path": "hf_space:results/single_episode_diagnostics/modality_ablation/ablation_metrics.csv",
+          "exists": true,
+          "bytes": 24871,
+          "sha256": "8e43bfa9b90544d7b09e207de26d1b852dd115b3e7c957eaa4b45e520dc050b8"
+        },
+        "hf_artifacts": {
+          "path": "hf_artifacts:results/single_episode_diagnostics/modality_ablation/ablation_metrics.csv",
+          "exists": true,
+          "bytes": 24871,
+          "sha256": "8e43bfa9b90544d7b09e207de26d1b852dd115b3e7c957eaa4b45e520dc050b8"
+        },
+        "hf_model": {
+          "path": "hf_model:results/single_episode_diagnostics/modality_ablation/ablation_metrics.csv",
+          "exists": true,
+          "bytes": 24871,
+          "sha256": "8e43bfa9b90544d7b09e207de26d1b852dd115b3e7c957eaa4b45e520dc050b8"
+        }
+      },
+      "failures": []
+    },
+    {
+      "name": "results/single_episode_diagnostics/modality_ablation/ablation_summary.json",
+      "status": "pass",
+      "local": {
+        "path": "repo:results/single_episode_diagnostics/modality_ablation/ablation_summary.json",
+        "exists": true,
+        "bytes": 735,
+        "sha256": "2761e1777963473ffa5110ab25e863a6b8ee19ce2004d94ffa938cfa5e0d93fc"
+      },
+      "mirrors": {
+        "hf_space": {
+          "path": "hf_space:results/single_episode_diagnostics/modality_ablation/ablation_summary.json",
+          "exists": true,
+          "bytes": 735,
+          "sha256": "2761e1777963473ffa5110ab25e863a6b8ee19ce2004d94ffa938cfa5e0d93fc"
+        },
+        "hf_artifacts": {
+          "path": "hf_artifacts:results/single_episode_diagnostics/modality_ablation/ablation_summary.json",
+          "exists": true,
+          "bytes": 735,
+          "sha256": "2761e1777963473ffa5110ab25e863a6b8ee19ce2004d94ffa938cfa5e0d93fc"
+        },
+        "hf_model": {
+          "path": "hf_model:results/single_episode_diagnostics/modality_ablation/ablation_summary.json",
+          "exists": true,
+          "bytes": 735,
+          "sha256": "2761e1777963473ffa5110ab25e863a6b8ee19ce2004d94ffa938cfa5e0d93fc"
+        }
+      },
+      "failures": []
+    },
+    {
+      "name": "results/single_episode_diagnostics/object_labels/object_vocab.json",
+      "status": "pass",
+      "local": {
+        "path": "repo:results/single_episode_diagnostics/object_labels/object_vocab.json",
+        "exists": true,
+        "bytes": 995,
+        "sha256": "813320bbe5d57c74803f5dfb080d440ef1804af41950e031d82ab80311eb04c7"
+      },
+      "mirrors": {
+        "hf_space": {
+          "path": "hf_space:results/single_episode_diagnostics/object_labels/object_vocab.json",
+          "exists": true,
+          "bytes": 995,
+          "sha256": "813320bbe5d57c74803f5dfb080d440ef1804af41950e031d82ab80311eb04c7"
+        },
+        "hf_artifacts": {
+          "path": "hf_artifacts:results/single_episode_diagnostics/object_labels/object_vocab.json",
+          "exists": true,
+          "bytes": 995,
+          "sha256": "813320bbe5d57c74803f5dfb080d440ef1804af41950e031d82ab80311eb04c7"
+        },
+        "hf_model": {
+          "path": "hf_model:results/single_episode_diagnostics/object_labels/object_vocab.json",
+          "exists": true,
+          "bytes": 995,
+          "sha256": "813320bbe5d57c74803f5dfb080d440ef1804af41950e031d82ab80311eb04c7"
+        }
+      },
+      "failures": []
+    },
+    {
+      "name": "results/single_episode_diagnostics/object_labels/window_object_labels.csv",
+      "status": "pass",
+      "local": {
+        "path": "repo:results/single_episode_diagnostics/object_labels/window_object_labels.csv",
+        "exists": true,
+        "bytes": 78160,
+        "sha256": "a28e43bf88f5c4c0a193b7c0c574526822e5ca0757d78b168c8450730d804510"
+      },
+      "mirrors": {
+        "hf_space": {
+          "path": "hf_space:results/single_episode_diagnostics/object_labels/window_object_labels.csv",
+          "exists": true,
+          "bytes": 78160,
+          "sha256": "a28e43bf88f5c4c0a193b7c0c574526822e5ca0757d78b168c8450730d804510"
+        },
+        "hf_artifacts": {
+          "path": "hf_artifacts:results/single_episode_diagnostics/object_labels/window_object_labels.csv",
+          "exists": true,
+          "bytes": 78160,
+          "sha256": "a28e43bf88f5c4c0a193b7c0c574526822e5ca0757d78b168c8450730d804510"
+        },
+        "hf_model": {
+          "path": "hf_model:results/single_episode_diagnostics/object_labels/window_object_labels.csv",
+          "exists": true,
+          "bytes": 78160,
+          "sha256": "a28e43bf88f5c4c0a193b7c0c574526822e5ca0757d78b168c8450730d804510"
+        }
+      },
+      "failures": []
+    },
+    {
+      "name": "results/single_episode_diagnostics/timeline_overlay/timeline_overlay.csv",
+      "status": "pass",
+      "local": {
+        "path": "repo:results/single_episode_diagnostics/timeline_overlay/timeline_overlay.csv",
+        "exists": true,
+        "bytes": 293713,
+        "sha256": "eebd53caf853aac5da51adea285e6623dd00860d23d3bc5703169ed6c53d0405"
+      },
+      "mirrors": {
+        "hf_space": {
+          "path": "hf_space:results/single_episode_diagnostics/timeline_overlay/timeline_overlay.csv",
+          "exists": true,
+          "bytes": 293713,
+          "sha256": "eebd53caf853aac5da51adea285e6623dd00860d23d3bc5703169ed6c53d0405"
+        },
+        "hf_artifacts": {
+          "path": "hf_artifacts:results/single_episode_diagnostics/timeline_overlay/timeline_overlay.csv",
+          "exists": true,
+          "bytes": 293713,
+          "sha256": "eebd53caf853aac5da51adea285e6623dd00860d23d3bc5703169ed6c53d0405"
+        },
+        "hf_model": {
+          "path": "hf_model:results/single_episode_diagnostics/timeline_overlay/timeline_overlay.csv",
+          "exists": true,
+          "bytes": 293713,
+          "sha256": "eebd53caf853aac5da51adea285e6623dd00860d23d3bc5703169ed6c53d0405"
+        }
+      },
+      "failures": []
+    },
+    {
+      "name": "results/single_episode_diagnostics/alignment_stress/alignment_shift_metrics.csv",
+      "status": "pass",
+      "local": {
+        "path": "repo:results/single_episode_diagnostics/alignment_stress/alignment_shift_metrics.csv",
+        "exists": true,
+        "bytes": 8203,
+        "sha256": "8b0026b472a0c0fd8c6eb5c9c307b41ac7ed71fe0ae3d52c8b18dfd5721f0ef1"
+      },
+      "mirrors": {
+        "hf_space": {
+          "path": "hf_space:results/single_episode_diagnostics/alignment_stress/alignment_shift_metrics.csv",
+          "exists": true,
+          "bytes": 8203,
+          "sha256": "8b0026b472a0c0fd8c6eb5c9c307b41ac7ed71fe0ae3d52c8b18dfd5721f0ef1"
+        },
+        "hf_artifacts": {
+          "path": "hf_artifacts:results/single_episode_diagnostics/alignment_stress/alignment_shift_metrics.csv",
+          "exists": true,
+          "bytes": 8203,
+          "sha256": "8b0026b472a0c0fd8c6eb5c9c307b41ac7ed71fe0ae3d52c8b18dfd5721f0ef1"
+        },
+        "hf_model": {
+          "path": "hf_model:results/single_episode_diagnostics/alignment_stress/alignment_shift_metrics.csv",
+          "exists": true,
+          "bytes": 8203,
+          "sha256": "8b0026b472a0c0fd8c6eb5c9c307b41ac7ed71fe0ae3d52c8b18dfd5721f0ef1"
+        }
+      },
+      "failures": []
+    },
+    {
+      "name": "results/single_episode_diagnostics/alignment_stress/alignment_stress_summary.json",
+      "status": "pass",
+      "local": {
+        "path": "repo:results/single_episode_diagnostics/alignment_stress/alignment_stress_summary.json",
+        "exists": true,
+        "bytes": 332,
+        "sha256": "2a467dae8e2f1606d77eccc8d1cbeb11507d462f95291d52b49d1057b7db9f14"
+      },
+      "mirrors": {
+        "hf_space": {
+          "path": "hf_space:results/single_episode_diagnostics/alignment_stress/alignment_stress_summary.json",
+          "exists": true,
+          "bytes": 332,
+          "sha256": "2a467dae8e2f1606d77eccc8d1cbeb11507d462f95291d52b49d1057b7db9f14"
+        },
+        "hf_artifacts": {
+          "path": "hf_artifacts:results/single_episode_diagnostics/alignment_stress/alignment_stress_summary.json",
+          "exists": true,
+          "bytes": 332,
+          "sha256": "2a467dae8e2f1606d77eccc8d1cbeb11507d462f95291d52b49d1057b7db9f14"
+        },
+        "hf_model": {
+          "path": "hf_model:results/single_episode_diagnostics/alignment_stress/alignment_stress_summary.json",
+          "exists": true,
+          "bytes": 332,
+          "sha256": "2a467dae8e2f1606d77eccc8d1cbeb11507d462f95291d52b49d1057b7db9f14"
+        }
+      },
+      "failures": []
+    },
     {
       "name": "docs/QUALITY_GATES.md",
       "status": "pass",

docs/data/publication_audit.json

@@ -1,6 +1,6 @@
 {
   "status": "pass",
-  "generated_at_utc": "2026-06-03T01:40:56+00:00",
+  "generated_at_utc": "2026-06-03T04:07:12+00:00",
   "checks": [
     {
       "name": "required_publication_assets_present",
@@ -179,8 +179,8 @@
     "github_repo": {
       "root": "repo",
       "exists": true,
-      "file_count": 333,
-      "text_file_count": 269,
+      "file_count": 352,
+      "text_file_count": 288,
       "largest_file": {
         "path": "results/episode_task_suite/modality_reconstruction/predictions.npz",
         "bytes": 52601010
@@ -190,19 +190,19 @@
     "hf_space_bundle": {
       "root": "hf_publish/space",
       "exists": true,
-      "file_count": 118,
-      "text_file_count": 91,
+      "file_count": 135,
+      "text_file_count": 108,
       "largest_file": {
-        "path": "assets/task_suite_infographic.png",
-        "bytes": 2618313
+        "path": "data/single_episode_explorer.json",
+        "bytes": 4101241
       },
       "violations": []
     },
     "hf_artifact_bundle": {
       "root": "hf_publish/artifacts",
       "exists": true,
-      "file_count": 362,
-      "text_file_count": 278,
+      "file_count": 381,
+      "text_file_count": 297,
       "largest_file": {
         "path": "results/episode_task_suite/modality_reconstruction/predictions.npz",
         "bytes": 52601010
@@ -212,8 +212,8 @@
     "hf_model_bundle": {
       "root": "hf_publish/model",
       "exists": true,
-      "file_count": 539,
-      "text_file_count": 423,
+      "file_count": 561,
+      "text_file_count": 445,
       "largest_file": {
         "path": "results/episode_task_suite/modality_reconstruction/predictions.npz",
         "bytes": 52601010

docs/data/website_integrity.json

@@ -1,14 +1,14 @@
 {
   "status": "pass",
-  "generated_at_utc": "2026-06-03T01:38:40+00:00",
+  "generated_at_utc": "2026-06-03T04:04:17+00:00",
   "docs_root": "docs",
   "site_base": "/ropedia-xperience-10m-task-suite/",
   "summary": {
-    "html_pages": 2,
-    "local_references": 100,
-    "external_reference_count": 80,
-    "json_files": 28,
-    "image_assets_referenced": 19,
+    "html_pages": 3,
+    "local_references": 109,
+    "external_reference_count": 83,
+    "json_files": 29,
+    "image_assets_referenced": 20,
     "failure_count": 0
   },
   "failures": {
@@ -74,8 +74,8 @@
       "name": "project_overview_precedes_progress_ledger",
       "status": "pass",
       "reason": "The project overview should appear before the deeper progress ledger.",
-      "overview_index": 60551,
-      "evidence_index": 73976
+      "overview_index": 60611,
+      "evidence_index": 74036
     },
     {
       "name": "project_status_links_json",
@@ -122,9 +122,9 @@
       "name": "evaluation_protocol_between_overview_and_progress",
       "status": "pass",
       "reason": "The evaluation protocol should appear before the deeper evidence ledger.",
-      "overview_index": 60551,
-      "protocol_index": 71379,
-      "evidence_index": 73976
+      "overview_index": 60611,
+      "protocol_index": 71439,
+      "evidence_index": 74036
     },
     {
       "name": "evaluation_protocol_links_json",
@@ -198,14 +198,20 @@
     {
       "path": "index.html",
       "id_count": 75,
-      "reference_count": 99,
+      "reference_count": 102,
       "image_count": 22
+    },
+    {
+      "path": "single_episode_explorer.html",
+      "id_count": 26,
+      "reference_count": 6,
+      "image_count": 1
     }
   ],
   "json_files": [
     {
       "path": "data/artifact_index.json",
-      "bytes": 27749,
+      "bytes": 29016,
       "top_level_type": "dict"
     },
     {
@@ -235,7 +241,7 @@
     },
     {
       "path": "data/mirror_parity.json",
-      "bytes": 76733,
+      "bytes": 81842,
       "top_level_type": "dict"
     },
     {
@@ -270,12 +276,12 @@
     },
     {
       "path": "data/publication_audit.json",
-      "bytes": 6962,
+      "bytes": 10473,
       "top_level_type": "dict"
     },
     {
       "path": "data/quality_gates.json",
-      "bytes": 7528,
+      "bytes": 8147,
       "top_level_type": "dict"
     },
     {
@@ -313,6 +319,11 @@
       "bytes": 20066,
       "top_level_type": "dict"
     },
+    {
+      "path": "data/single_episode_explorer.json",
+      "bytes": 4101241,
+      "top_level_type": "dict"
+    },
     {
       "path": "data/source_alignment_audit.json",
       "bytes": 4432,
@@ -335,7 +346,7 @@
     },
     {
       "path": "data/website_integrity.json",
-      "bytes": 12406,
+      "bytes": 13148,
       "top_level_type": "dict"
     },
     {
@@ -353,6 +364,14 @@
       "height": 64,
       "format": "PNG"
     },
+    {
+      "path": "assets/brand/xperience10m-logo-mark-192.png",
+      "exists": true,
+      "bytes": 41318,
+      "width": 192,
+      "height": 192,
+      "format": "PNG"
+    },
     {
       "path": "assets/charts/cross_modal_retrieval.svg",
       "exists": true,

docs/index.html

@@ -1880,6 +1880,7 @@
         <a href="#models">Results</a>
         <a href="#directions">Directions</a>
         <a href="#walkthroughs">Walkthrough</a>
+        <a href="single_episode_explorer.html">Explorer</a>
         <a href="#artifacts">Resources</a>
         <a class="nav-action" href="https://huggingface.co/spaces/cy0307/ropedia-xperience-10m-task-suite">HF Space</a>
         <a class="nav-action" href="https://github.com/ChaoYue0307/ropedia-xperience-10m-task-suite">GitHub</a>
@@ -2711,6 +2712,7 @@
           <img class="chart" src="assets/charts/episode_task_scores_neural_mlp.svg" alt="Neural MLP task score chart">
           <img class="chart" src="assets/charts/episode_task_scores_minimal_vs_neural.svg" alt="Minimal versus neural score chart">
         </div>
+        <p class="section-note"><a href="single_episode_explorer.html">Open the single-episode explorer</a> to inspect window-level labels, predictions, feature-block statistics, object labels, and diagnostic scores.</p>
       </div>
     </section>
 
@@ -2752,6 +2754,7 @@
               <article class="artifact"><h3>Four-direction taxonomy</h3><p>Generated JSON, CSV, Markdown, and website data mapping all 12 tasks to the four research tracks.</p><a href="https://github.com/ChaoYue0307/ropedia-xperience-10m-task-suite/tree/main/results/episode_task_suite/research_directions">research_directions/</a></article>
               <article class="artifact"><h3>Direction extension probes</h3><p>Four coded probes, one per research direction, with minimal and neural metrics plus prediction/rank CSVs.</p><a href="https://github.com/ChaoYue0307/ropedia-xperience-10m-task-suite/tree/main/results/episode_task_suite/research_direction_extensions">research_direction_extensions/</a></article>
               <article class="artifact"><h3>Task walkthroughs</h3><p>Case studies for all 12 tasks, including input, middle process modules, output, metric, limitation, and task-player data.</p><a href="https://github.com/ChaoYue0307/ropedia-xperience-10m-task-suite/tree/main/results/episode_task_suite/task_walkthroughs">task_walkthroughs/</a></article>
+              <article class="artifact"><h3>Single-episode explorer</h3><p>Interactive window-level view of labels, predictions, feature-block statistics, object labels, and diagnostics.</p><a href="single_episode_explorer.html">single_episode_explorer.html</a></article>
               <article class="artifact"><h3>Cross-modal retrieval</h3><p>The strongest self-supervised signal from the single episode.</p><a href="https://github.com/ChaoYue0307/ropedia-xperience-10m-task-suite/blob/main/results/episode_task_suite/cross_modal_retrieval/metrics.json">metrics.json</a></article>
             </div>
           </section>

results/single_episode_diagnostics/README.md

@@ -0,0 +1,23 @@
+# Single-Episode Diagnostics Index
+
+These outputs are local diagnostics built from the existing one-episode Xperience-10M artifacts. They are designed for manual verification while waiting for full multi-episode data access.
+
+## Generated Analyses
+
+- `modality_ablation/`: compact ridge-head ablations across real feature blocks.
+- `timeline_overlay/`: existing prediction CSVs aligned to the episode timeline.
+- `alignment_stress/`: cross-modal retrieval under explicit time-shift perturbations.
+- `provenance.json`: input hashes, feature dimensions, and source artifact identifiers.
+
+## Validity Boundaries
+
+- This is a single-episode diagnostic, not a full Xperience-10M benchmark.
+- Rows marked `not_computed` are intentionally left blank when train labels or valid splits are unavailable.
+- Rows marked `derived_perturbation` use real features with deliberate time shifts for stress testing.
+
+## Counts
+
+- Ablation rows: 96; computed: 96.
+- Timeline overlay rows: 2079.
+- Alignment stress rows: 45.
+- Shared feature shape: 1161 windows x 8378 features.

results/single_episode_diagnostics/alignment_stress/ALIGNMENT_STRESS_REPORT.md

@@ -0,0 +1,17 @@
+# Cross-Modal Alignment Stress Report
+
+This diagnostic uses real held-out feature windows, then deliberately shifts the query modality in time at evaluation. The perturbation is derived; it is not treated as observed data.
+
+## Zero-Shift Versus Worst Shift
+
+- Inertial: zero-shift MRR=0.2840; worst shift=-20 windows, MRR=0.0199
+- Language: zero-shift MRR=0.0310; worst shift=-40 windows, MRR=0.0158
+- Motion Capture: zero-shift MRR=0.2553; worst shift=-10 windows, MRR=0.0183
+- Motion + Pose + IMU: zero-shift MRR=0.3897; worst shift=-20 windows, MRR=0.0238
+- Pose + SLAM: zero-shift MRR=0.4262; worst shift=-20 windows, MRR=0.0206
+
+## Files
+
+- `alignment_shift_metrics.csv`: MRR/rank metrics for each query group and time shift.
+- `alignment_shift_curves.svg`: MRR curves across time shifts.
+- `alignment_stress_summary.json`: perturbation definition and status.

results/single_episode_diagnostics/alignment_stress/alignment_shift_metrics.csv

@@ -0,0 +1,46 @@
+query_group,query_display,target_group,shift_windows,shift_frames,status,mrr,top1_accuracy,top5_accuracy,top10_accuracy,median_rank,mean_rank,num_queries
+motion_capture,Motion Capture,depth_plus_video,-40,-200,derived_perturbation,0.02984776347875595,0.006493506493506494,0.032467532467532464,0.048701298701298704,119.0,125.72402954101562,308
+motion_capture,Motion Capture,depth_plus_video,-20,-100,derived_perturbation,0.019169922918081284,0.003048780487804878,0.003048780487804878,0.018292682926829267,153.0,154.24085998535156,328
+motion_capture,Motion Capture,depth_plus_video,-10,-50,derived_perturbation,0.01829293556511402,0.0,0.0029585798816568047,0.023668639053254437,106.0,130.5325469970703,338
+motion_capture,Motion Capture,depth_plus_video,-5,-25,derived_perturbation,0.028951412066817284,0.0,0.014577259475218658,0.05830903790087463,67.0,89.93585968017578,343
+motion_capture,Motion Capture,depth_plus_video,0,0,derived_perturbation,0.2553335726261139,0.15804597701149425,0.35344827586206895,0.3994252873563218,21.5,49.181034088134766,348
+motion_capture,Motion Capture,depth_plus_video,5,25,derived_perturbation,0.04436318948864937,0.008746355685131196,0.037900874635568516,0.08746355685131195,64.0,83.76384735107422,343
+motion_capture,Motion Capture,depth_plus_video,10,50,derived_perturbation,0.026273079216480255,0.0,0.014792899408284023,0.047337278106508875,77.0,106.73668670654297,338
+motion_capture,Motion Capture,depth_plus_video,20,100,derived_perturbation,0.023496314883232117,0.003048780487804878,0.018292682926829267,0.04573170731707317,108.5,137.9176788330078,328
+motion_capture,Motion Capture,depth_plus_video,40,200,derived_perturbation,0.02917252667248249,0.006493506493506494,0.025974025974025976,0.05519480519480519,110.5,121.33441925048828,308
+pose_slam,Pose + SLAM,depth_plus_video,-40,-200,derived_perturbation,0.03615332394838333,0.003246753246753247,0.03571428571428571,0.07467532467532467,98.0,114.43506622314453,308
+pose_slam,Pose + SLAM,depth_plus_video,-20,-100,derived_perturbation,0.02059117704629898,0.003048780487804878,0.012195121951219513,0.024390243902439025,109.5,137.0731658935547,328
+pose_slam,Pose + SLAM,depth_plus_video,-10,-50,derived_perturbation,0.04128313437104225,0.005917159763313609,0.038461538461538464,0.07692307692307693,72.0,103.94674682617188,338
+pose_slam,Pose + SLAM,depth_plus_video,-5,-25,derived_perturbation,0.05835483595728874,0.011661807580174927,0.061224489795918366,0.119533527696793,43.0,58.218658447265625,343
+pose_slam,Pose + SLAM,depth_plus_video,0,0,derived_perturbation,0.42622581124305725,0.3017241379310345,0.5488505747126436,0.6551724137931034,4.0,15.623562812805176,348
+pose_slam,Pose + SLAM,depth_plus_video,5,25,derived_perturbation,0.04654298722743988,0.0058309037900874635,0.04956268221574344,0.11661807580174927,55.0,66.43148803710938,343
+pose_slam,Pose + SLAM,depth_plus_video,10,50,derived_perturbation,0.034309279173612595,0.005917159763313609,0.023668639053254437,0.05621301775147929,71.0,100.44082641601562,338
+pose_slam,Pose + SLAM,depth_plus_video,20,100,derived_perturbation,0.03287472575902939,0.006097560975609756,0.03048780487804878,0.06097560975609756,97.5,127.41158294677734,328
+pose_slam,Pose + SLAM,depth_plus_video,40,200,derived_perturbation,0.024975253269076347,0.003246753246753247,0.016233766233766232,0.03571428571428571,88.5,116.36363983154297,308
+inertial,Inertial,depth_plus_video,-40,-200,derived_perturbation,0.042965441942214966,0.00974025974025974,0.045454545454545456,0.09090909090909091,90.0,116.86363983154297,308
+inertial,Inertial,depth_plus_video,-20,-100,derived_perturbation,0.019861916080117226,0.003048780487804878,0.009146341463414634,0.01524390243902439,112.0,135.9573211669922,328
+inertial,Inertial,depth_plus_video,-10,-50,derived_perturbation,0.04950016736984253,0.011834319526627219,0.05325443786982249,0.10946745562130178,74.0,102.37574005126953,338
+inertial,Inertial,depth_plus_video,-5,-25,derived_perturbation,0.05499911680817604,0.0029154518950437317,0.05830903790087463,0.13994169096209913,39.0,63.70845413208008,343
+inertial,Inertial,depth_plus_video,0,0,derived_perturbation,0.2840072810649872,0.16379310344827586,0.3735632183908046,0.5229885057471264,10.0,20.577587127685547,348
+inertial,Inertial,depth_plus_video,5,25,derived_perturbation,0.054161082953214645,0.014577259475218658,0.04956268221574344,0.10495626822157435,53.0,68.86006164550781,343
+inertial,Inertial,depth_plus_video,10,50,derived_perturbation,0.03178250044584274,0.005917159763313609,0.008875739644970414,0.05917159763313609,76.0,103.02071380615234,338
+inertial,Inertial,depth_plus_video,20,100,derived_perturbation,0.03213934600353241,0.009146341463414634,0.021341463414634148,0.042682926829268296,93.5,125.67987823486328,328
+inertial,Inertial,depth_plus_video,40,200,derived_perturbation,0.031400587409734726,0.003246753246753247,0.03896103896103896,0.05194805194805195,91.0,115.68830871582031,308
+language,Language,depth_plus_video,-40,-200,derived_perturbation,0.015811588615179062,0.0,0.003246753246753247,0.016233766233766232,145.5,141.49026489257812,308
+language,Language,depth_plus_video,-20,-100,derived_perturbation,0.027325116097927094,0.006097560975609756,0.024390243902439025,0.06097560975609756,174.0,162.0792694091797,328
+language,Language,depth_plus_video,-10,-50,derived_perturbation,0.02521640993654728,0.0029585798816568047,0.023668639053254437,0.05621301775147929,165.0,162.10354614257812,338
+language,Language,depth_plus_video,-5,-25,derived_perturbation,0.02469729632139206,0.0029154518950437317,0.02040816326530612,0.04956268221574344,165.0,158.99708557128906,343
+language,Language,depth_plus_video,0,0,derived_perturbation,0.031006580218672752,0.005747126436781609,0.031609195402298854,0.05747126436781609,138.0,146.83045959472656,348
+language,Language,depth_plus_video,5,25,derived_perturbation,0.04090346768498421,0.008746355685131196,0.04956268221574344,0.08454810495626822,102.0,135.07289123535156,343
+language,Language,depth_plus_video,10,50,derived_perturbation,0.0362100675702095,0.008875739644970414,0.03254437869822485,0.07692307692307693,101.5,131.18934631347656,338
+language,Language,depth_plus_video,20,100,derived_perturbation,0.03773954510688782,0.009146341463414634,0.036585365853658534,0.08231707317073171,111.0,137.8353729248047,328
+language,Language,depth_plus_video,40,200,derived_perturbation,0.037675727158784866,0.00974025974025974,0.04220779220779221,0.08116883116883117,139.5,144.4837646484375,308
+motion_pose_inertial,Motion + Pose + IMU,depth_plus_video,-40,-200,derived_perturbation,0.05048111826181412,0.016233766233766232,0.05519480519480519,0.1038961038961039,99.5,116.717529296875,308
+motion_pose_inertial,Motion + Pose + IMU,depth_plus_video,-20,-100,derived_perturbation,0.023761091753840446,0.003048780487804878,0.01524390243902439,0.021341463414634148,111.0,140.868896484375,328
+motion_pose_inertial,Motion + Pose + IMU,depth_plus_video,-10,-50,derived_perturbation,0.039904821664094925,0.008875739644970414,0.03550295857988166,0.07988165680473373,81.5,107.72189331054688,338
+motion_pose_inertial,Motion + Pose + IMU,depth_plus_video,-5,-25,derived_perturbation,0.051686227321624756,0.008746355685131196,0.043731778425655975,0.12244897959183673,47.0,63.40524673461914,343
+motion_pose_inertial,Motion + Pose + IMU,depth_plus_video,0,0,derived_perturbation,0.38971078395843506,0.28448275862068967,0.4827586206896552,0.5718390804597702,6.0,25.27011489868164,348
+motion_pose_inertial,Motion + Pose + IMU,depth_plus_video,5,25,derived_perturbation,0.05908266454935074,0.014577259475218658,0.0641399416909621,0.13119533527696792,48.0,64.80757904052734,343
+motion_pose_inertial,Motion + Pose + IMU,depth_plus_video,10,50,derived_perturbation,0.03273069113492966,0.0029585798816568047,0.01775147928994083,0.06804733727810651,63.0,96.31952667236328,338
+motion_pose_inertial,Motion + Pose + IMU,depth_plus_video,20,100,derived_perturbation,0.028844518586993217,0.006097560975609756,0.024390243902439025,0.04573170731707317,81.5,131.6280517578125,328
+motion_pose_inertial,Motion + Pose + IMU,depth_plus_video,40,200,derived_perturbation,0.033500298857688904,0.00974025974025974,0.032467532467532464,0.048701298701298704,99.0,120.70130157470703,308

results/single_episode_diagnostics/alignment_stress/alignment_stress_summary.json

@@ -0,0 +1,6 @@
+{
+  "description": "Real feature windows are deliberately time-shifted at evaluation time to test cross-modal alignment sensitivity.",
+  "target_group": "depth_confidence + video_*",
+  "status_meaning": "derived_perturbation means the features are real but the time shift is an explicit diagnostic perturbation.",
+  "num_rows": 45
+}

results/single_episode_diagnostics/modality_ablation/MODALITY_ABLATION_REPORT.md

@@ -0,0 +1,26 @@
+# Single-Episode Modality Ablation Report
+
+This diagnostic reruns compact ridge heads on the exported one-episode feature matrix. It is useful for checking which real feature blocks can support each task on this episode, not for estimating dataset-wide generalization.
+
+No synthetic labels are introduced. Derived proxy targets are marked in `target_variant`, and feature groups that overlap with the target source are marked in `target_source_overlap`.
+
+## Best Computed Group Per Task
+
+- Current Action Recognition: Language score=0.0278, macro_f1=0.0278, target overlap=false
+- Current Subtask Recognition: Language score=0.0483, macro_f1=0.0483, target overlap=false
+- Action Transition Detection: Language score=0.7052, macro_f1=0.7052, target overlap=false
+- Next-Action Prediction: Language score=0.0419, macro_f1=0.0419, target overlap=false
+- Future Hand Motion Forecasting: Inertial score=0.5679, mae=0.7608, target overlap=false
+- Contact State Prediction: All Features score=1.0000, macro_f1=1.0000, target overlap=false
+- Relevant Object Prediction: Language score=0.2302, micro_f1=0.2302, target overlap=true; best non-overlap: Depth score=0.2013, micro_f1=0.2013
+- Language-to-Time Grounding: Language score=0.2453, mrr=0.2453, target overlap=true; best non-overlap: Motion Capture score=0.0306, mrr=0.0306
+- Cross-Modal Window Retrieval: All Features score=0.9724, mrr=0.9724, target overlap=true; best non-overlap: Pose + SLAM score=0.4262, mrr=0.4262
+- Sensor-to-Visual Reconstruction: Video score=0.6113, mae=0.6358, target overlap=true; best non-overlap: Pose + SLAM score=0.5359, mae=0.8659
+- Temporal Order Verification: Pose + SLAM score=0.5259, macro_f1=0.5259, target overlap=false
+- Cross-Modal Misalignment Detection: Video score=0.4949, macro_f1=0.4949, target overlap=false
+
+## Files
+
+- `ablation_metrics.csv`: every task/modality pair, including not-computed rows and reasons.
+- `ablation_matrix.svg`: compact heatmap for manual inspection.
+- `ablation_summary.json`: group dimensions and computed/not-computed counts.

results/single_episode_diagnostics/modality_ablation/ablation_metrics.csv

@@ -0,0 +1,97 @@
+task,task_display,modality_group,modality_display,status,score,primary_metric,primary_metric_value,target_variant,target_source_overlap,reason,accuracy,macro_f1,balanced_accuracy,num_classes,num_train,num_test,unseen_test_classes,unseen_test_class_count,mse,mae,r2,micro_f1,exact_match,precision,recall,num_objects,mrr,top1_accuracy,top5_accuracy,top10_accuracy,median_rank,mean_rank,num_queries
+timeline_action,Current Action Recognition,all_features,All Features,computed,0.008771929824561405,macro_f1,0.008771929824561405,,false,,0.020114942528735632,0.008771929824561405,0.005668016194331984,19,813,348,Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee,4,,,,,,,,,,,,,,,
+timeline_action,Current Action Recognition,video,Video,computed,0.0066280033140016575,macro_f1,0.0066280033140016575,,false,,0.011494252873563218,0.0066280033140016575,0.0036199095022624436,19,813,348,Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee,4,,,,,,,,,,,,,,,
+timeline_action,Current Action Recognition,depth,Depth,computed,0.0030075187969924814,macro_f1,0.0030075187969924814,,false,,0.005747126436781609,0.0030075187969924814,0.001619433198380567,19,813,348,Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee,4,,,,,,,,,,,,,,,
+timeline_action,Current Action Recognition,pose_slam,Pose + SLAM,computed,0.0,macro_f1,0.0,,false,,0.0,0.0,0.0,19,813,348,Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee,4,,,,,,,,,,,,,,,
+timeline_action,Current Action Recognition,motion_capture,Motion Capture,computed,0.0055147058823529415,macro_f1,0.0055147058823529415,,false,,0.008620689655172414,0.0055147058823529415,0.0028846153846153848,19,813,348,Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee,4,,,,,,,,,,,,,,,
+timeline_action,Current Action Recognition,inertial,Inertial,computed,0.003055767761650115,macro_f1,0.003055767761650115,,false,,0.005747126436781609,0.003055767761650115,0.0018099547511312218,19,813,348,Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee,4,,,,,,,,,,,,,,,
+timeline_action,Current Action Recognition,language,Language,computed,0.027777777777777776,macro_f1,0.027777777777777776,,false,,0.05747126436781609,0.027777777777777776,0.03615384615384616,19,813,348,Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee,4,,,,,,,,,,,,,,,
+timeline_action,Current Action Recognition,no_language,All Except Language,computed,0.007112375533428165,macro_f1,0.007112375533428165,,false,,0.014367816091954023,0.007112375533428165,0.004048582995951417,19,813,348,Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee,4,,,,,,,,,,,,,,,
+timeline_subtask,Current Subtask Recognition,all_features,All Features,computed,0.0111731843575419,macro_f1,0.0111731843575419,,false,,0.040229885057471264,0.0111731843575419,0.017543859649122806,15,813,348,Move bottle to coffee equipment|Prepare for pouring|Pour coffee|Pour milk into coffee,4,,,,,,,,,,,,,,,
+timeline_subtask,Current Subtask Recognition,video,Video,computed,0.011740041928721174,macro_f1,0.011740041928721174,,false,,0.040229885057471264,0.011740041928721174,0.01637426900584795,15,813,348,Move bottle to coffee equipment|Prepare for pouring|Pour coffee|Pour milk into coffee,4,,,,,,,,,,,,,,,
+timeline_subtask,Current Subtask Recognition,depth,Depth,computed,0.009467455621301775,macro_f1,0.009467455621301775,,false,,0.022988505747126436,0.009467455621301775,0.010796221322537112,15,813,348,Move bottle to coffee equipment|Prepare for pouring|Pour coffee|Pour milk into coffee,4,,,,,,,,,,,,,,,
+timeline_subtask,Current Subtask Recognition,pose_slam,Pose + SLAM,computed,0.002331002331002331,macro_f1,0.002331002331002331,,false,,0.0028735632183908046,0.002331002331002331,0.001349527665317139,15,813,348,Move bottle to coffee equipment|Prepare for pouring|Pour coffee|Pour milk into coffee,4,,,,,,,,,,,,,,,
+timeline_subtask,Current Subtask Recognition,motion_capture,Motion Capture,computed,0.006756756756756756,macro_f1,0.006756756756756756,,false,,0.008620689655172414,0.006756756756756756,0.0043859649122807015,15,813,348,Move bottle to coffee equipment|Prepare for pouring|Pour coffee|Pour milk into coffee,4,,,,,,,,,,,,,,,
+timeline_subtask,Current Subtask Recognition,inertial,Inertial,computed,0.004662004662004662,macro_f1,0.004662004662004662,,false,,0.005747126436781609,0.004662004662004662,0.002699055330634278,15,813,348,Move bottle to coffee equipment|Prepare for pouring|Pour coffee|Pour milk into coffee,4,,,,,,,,,,,,,,,
+timeline_subtask,Current Subtask Recognition,language,Language,computed,0.04828150572831424,macro_f1,0.04828150572831424,,false,,0.14655172413793102,0.04828150572831424,0.0939327485380117,15,813,348,Move bottle to coffee equipment|Prepare for pouring|Pour coffee|Pour milk into coffee,4,,,,,,,,,,,,,,,
+timeline_subtask,Current Subtask Recognition,no_language,All Except Language,computed,0.012658227848101266,macro_f1,0.012658227848101266,,false,,0.03735632183908046,0.012658227848101266,0.017543859649122806,15,813,348,Move bottle to coffee equipment|Prepare for pouring|Pour coffee|Pour milk into coffee,4,,,,,,,,,,,,,,,
+transition_detection,Action Transition Detection,all_features,All Features,computed,0.46870229007633585,macro_f1,0.46870229007633585,,false,,0.882183908045977,0.46870229007633585,0.4623493975903614,2,813,348,,0,,,,,,,,,,,,,,,
+transition_detection,Action Transition Detection,video,Video,computed,0.46625766871165636,macro_f1,0.46625766871165636,,false,,0.8735632183908046,0.46625766871165636,0.4578313253012048,2,813,348,,0,,,,,,,,,,,,,,,
+transition_detection,Action Transition Detection,depth,Depth,computed,0.4604651162790698,macro_f1,0.4604651162790698,,false,,0.853448275862069,0.4604651162790698,0.44728915662650603,2,813,348,,0,,,,,,,,,,,,,,,
+transition_detection,Action Transition Detection,pose_slam,Pose + SLAM,computed,0.48444444444444446,macro_f1,0.48444444444444446,,false,,0.9396551724137931,0.48444444444444446,0.4924698795180723,2,813,348,,0,,,,,,,,,,,,,,,
+transition_detection,Action Transition Detection,motion_capture,Motion Capture,computed,0.5439056356487549,macro_f1,0.5439056356487549,,false,,0.896551724137931,0.5439056356487549,0.5591114457831325,2,813,348,,0,,,,,,,,,,,,,,,
+transition_detection,Action Transition Detection,inertial,Inertial,computed,0.48520710059171596,macro_f1,0.48520710059171596,,false,,0.9425287356321839,0.48520710059171596,0.4939759036144578,2,813,348,,0,,,,,,,,,,,,,,,
+transition_detection,Action Transition Detection,language,Language,computed,0.7051957831325302,macro_f1,0.7051957831325302,,false,,0.9482758620689655,0.7051957831325302,0.7051957831325302,2,813,348,,0,,,,,,,,,,,,,,,
+transition_detection,Action Transition Detection,no_language,All Except Language,computed,0.46543778801843316,macro_f1,0.46543778801843316,,false,,0.8706896551724138,0.46543778801843316,0.4563253012048193,2,813,348,,0,,,,,,,,,,,,,,,
+next_action,Next-Action Prediction,all_features,All Features,computed,0.0060882800608828,macro_f1,0.0060882800608828,future action label from windows.csv,false,,0.011527377521613832,0.0060882800608828,0.003472222222222222,19,810,347,Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee,4,,,,,,,,,,,,,,,
+next_action,Next-Action Prediction,video,Video,computed,0.006349206349206349,macro_f1,0.006349206349206349,future action label from windows.csv,false,,0.011527377521613832,0.006349206349206349,0.003472222222222222,19,810,347,Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee,4,,,,,,,,,,,,,,,
+next_action,Next-Action Prediction,depth,Depth,computed,0.001594896331738437,macro_f1,0.001594896331738437,future action label from windows.csv,false,,0.002881844380403458,0.001594896331738437,0.0008223684210526315,19,810,347,Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee,4,,,,,,,,,,,,,,,
+next_action,Next-Action Prediction,pose_slam,Pose + SLAM,computed,0.0,macro_f1,0.0,future action label from windows.csv,false,,0.0,0.0,0.0,19,810,347,Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee,4,,,,,,,,,,,,,,,
+next_action,Next-Action Prediction,motion_capture,Motion Capture,computed,0.00322061191626409,macro_f1,0.00322061191626409,future action label from windows.csv,false,,0.005763688760806916,0.00322061191626409,0.001736111111111111,19,810,347,Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee,4,,,,,,,,,,,,,,,
+next_action,Next-Action Prediction,inertial,Inertial,computed,0.00196078431372549,macro_f1,0.00196078431372549,future action label from windows.csv,false,,0.002881844380403458,0.00196078431372549,0.0010416666666666667,19,810,347,Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee,4,,,,,,,,,,,,,,,
+next_action,Next-Action Prediction,language,Language,computed,0.04193971166448231,macro_f1,0.04193971166448231,future action label from windows.csv,false,,0.1844380403458213,0.04193971166448231,0.07142857142857142,19,810,347,Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee,4,,,,,,,,,,,,,,,
+next_action,Next-Action Prediction,no_language,All Except Language,computed,0.004511278195488722,macro_f1,0.004511278195488722,future action label from windows.csv,false,,0.008645533141210375,0.004511278195488722,0.0024671052631578946,19,810,347,Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee,4,,,,,,,,,,,,,,,
+hand_trajectory_forecast,Future Hand Motion Forecasting,all_features,All Features,computed,0.1047945346490482,mae,8.542482376098633,future hand feature vector from shared_windows.npz,true,,,,,,,,,,6413.8505859375,8.542482376098633,-6684.484259411514,,,,,,,,,,,,
+hand_trajectory_forecast,Future Hand Motion Forecasting,video,Video,computed,0.4956350584748486,mae,1.0176135301589966,future hand feature vector from shared_windows.npz,false,,,,,,,,,,1.7896661758422852,1.0176135301589966,-0.8654605965108897,,,,,,,,,,,,
+hand_trajectory_forecast,Future Hand Motion Forecasting,depth,Depth,computed,0.04014931629731973,mae,23.907024383544922,future hand feature vector from shared_windows.npz,false,,,,,,,,,,72553.34375,23.907024383544922,-75625.0610993949,,,,,,,,,,,,
+hand_trajectory_forecast,Future Hand Motion Forecasting,pose_slam,Pose + SLAM,computed,0.5611809661721311,mae,0.7819563746452332,future hand feature vector from shared_windows.npz,false,,,,,,,,,,1.2600995302200317,0.7819563746452332,-0.3134661692106211,,,,,,,,,,,,
+hand_trajectory_forecast,Future Hand Motion Forecasting,motion_capture,Motion Capture,computed,0.0839705207556719,mae,10.908941268920898,future hand feature vector from shared_windows.npz,true,,,,,,,,,,6293.8876953125,10.908941268920898,-6559.441194341517,,,,,,,,,,,,
+hand_trajectory_forecast,Future Hand Motion Forecasting,inertial,Inertial,computed,0.5679183061202404,mae,0.7608166337013245,future hand feature vector from shared_windows.npz,false,,,,,,,,,,1.1916581392288208,0.7608166337013245,-0.24212624676650907,,,,,,,,,,,,
+hand_trajectory_forecast,Future Hand Motion Forecasting,language,Language,computed,0.451525705011023,mae,1.2147133350372314,future hand feature vector from shared_windows.npz,false,,,,,,,,,,2.3450045585632324,1.2147133350372314,-1.4443180759924243,,,,,,,,,,,,
+hand_trajectory_forecast,Future Hand Motion Forecasting,no_language,All Except Language,computed,0.09737268805379895,mae,9.269820213317871,future hand feature vector from shared_windows.npz,true,,,,,,,,,,7166.751953125,9.269820213317871,-7469.272088447983,,,,,,,,,,,,
+contact_prediction,Contact State Prediction,all_features,All Features,computed,1.0,macro_f1,1.0,contact proxy derived from body_contacts feature block,false,,1.0,1.0,1.0,1,813,348,,0,,,,,,,,,,,,,,,
+contact_prediction,Contact State Prediction,video,Video,computed,1.0,macro_f1,1.0,contact proxy derived from body_contacts feature block,false,,1.0,1.0,1.0,1,813,348,,0,,,,,,,,,,,,,,,
+contact_prediction,Contact State Prediction,depth,Depth,computed,1.0,macro_f1,1.0,contact proxy derived from body_contacts feature block,false,,1.0,1.0,1.0,1,813,348,,0,,,,,,,,,,,,,,,
+contact_prediction,Contact State Prediction,pose_slam,Pose + SLAM,computed,1.0,macro_f1,1.0,contact proxy derived from body_contacts feature block,false,,1.0,1.0,1.0,1,813,348,,0,,,,,,,,,,,,,,,
+contact_prediction,Contact State Prediction,motion_capture,Motion Capture,computed,1.0,macro_f1,1.0,contact proxy derived from body_contacts feature block,false,,1.0,1.0,1.0,1,813,348,,0,,,,,,,,,,,,,,,
+contact_prediction,Contact State Prediction,inertial,Inertial,computed,1.0,macro_f1,1.0,contact proxy derived from body_contacts feature block,false,,1.0,1.0,1.0,1,813,348,,0,,,,,,,,,,,,,,,
+contact_prediction,Contact State Prediction,language,Language,computed,1.0,macro_f1,1.0,contact proxy derived from body_contacts feature block,false,,1.0,1.0,1.0,1,813,348,,0,,,,,,,,,,,,,,,
+contact_prediction,Contact State Prediction,no_language,All Except Language,computed,1.0,macro_f1,1.0,contact proxy derived from body_contacts feature block,false,,1.0,1.0,1.0,1,813,348,,0,,,,,,,,,,,,,,,
+object_relevance,Relevant Object Prediction,all_features,All Features,computed,0.175914508836827,micro_f1,0.175914508836827,object sets exported from annotation.hdf5 caption_frame_info_map,true,,,0.06322379109578449,,,813,348,,,,,,0.175914508836827,0.020114942528735632,0.19888475836431227,0.15770081061164334,34,,,,,,,
+object_relevance,Relevant Object Prediction,video,Video,computed,0.14804270462633454,micro_f1,0.14804270462633454,object sets exported from annotation.hdf5 caption_frame_info_map,false,,,0.04379950367755125,,,813,348,,,,,,0.14804270462633454,0.008620689655172414,0.14315209910529939,0.15327929255711129,34,,,,,,,
+object_relevance,Relevant Object Prediction,depth,Depth,computed,0.20134228187919462,micro_f1,0.20134228187919462,object sets exported from annotation.hdf5 caption_frame_info_map,false,,,0.0649677953734521,,,813,348,,,,,,0.20134228187919462,0.011494252873563218,0.18484288354898337,0.2210759027266028,34,,,,,,,
+object_relevance,Relevant Object Prediction,pose_slam,Pose + SLAM,computed,0.19528071602929212,micro_f1,0.19528071602929212,object sets exported from annotation.hdf5 caption_frame_info_map,false,,,0.05592381693865655,,,813,348,,,,,,0.19528071602929212,0.0,0.21798365122615804,0.17686072218128224,34,,,,,,,
+object_relevance,Relevant Object Prediction,motion_capture,Motion Capture,computed,0.11607786589762077,micro_f1,0.11607786589762077,object sets exported from annotation.hdf5 caption_frame_info_map,false,,,0.045395437036303915,,,813,348,,,,,,0.11607786589762077,0.0028735632183908046,0.11362032462949895,0.11864406779661017,34,,,,,,,
+object_relevance,Relevant Object Prediction,inertial,Inertial,computed,0.1716082659478886,micro_f1,0.1716082659478886,object sets exported from annotation.hdf5 caption_frame_info_map,false,,,0.04806995854957751,,,813,348,,,,,,0.1716082659478886,0.0,0.21979286536248563,0.14075165806927045,34,,,,,,,
+object_relevance,Relevant Object Prediction,language,Language,computed,0.23021032504780117,micro_f1,0.23021032504780117,object sets exported from annotation.hdf5 caption_frame_info_map,true,,,0.0947530205484707,,,813,348,,,,,,0.23021032504780117,0.15229885057471265,0.23926868044515104,0.22181282240235814,34,,,,,,,
+object_relevance,Relevant Object Prediction,no_language,All Except Language,computed,0.14793328498912256,micro_f1,0.14793328498912256,object sets exported from annotation.hdf5 caption_frame_info_map,false,,,0.05137956064750565,,,813,348,,,,,,0.14793328498912256,0.008620689655172414,0.145610278372591,0.1503316138540899,34,,,,,,,
+caption_grounding,Language-to-Time Grounding,all_features,All Features,computed,0.21027426421642303,mrr,0.21027426421642303,,true,,,,,,,,,,,,,,,,,,0.21027426421642303,0.08908045977011494,0.33045977011494254,0.4482758620689655,13.0,22.55172348022461,348
+caption_grounding,Language-to-Time Grounding,video,Video,computed,0.022670436650514603,mrr,0.022670436650514603,,false,,,,,,,,,,,,,,,,,,0.022670436650514603,0.0028735632183908046,0.02586206896551724,0.034482758620689655,162.0,161.4770050048828,348
+caption_grounding,Language-to-Time Grounding,depth,Depth,computed,0.02443847246468067,mrr,0.02443847246468067,,false,,,,,,,,,,,,,,,,,,0.02443847246468067,0.0028735632183908046,0.020114942528735632,0.03735632183908046,114.0,137.90805053710938,348
+caption_grounding,Language-to-Time Grounding,pose_slam,Pose + SLAM,computed,0.02946249581873417,mrr,0.02946249581873417,,false,,,,,,,,,,,,,,,,,,0.02946249581873417,0.008620689655172414,0.028735632183908046,0.04597701149425287,143.5,155.4712677001953,348
+caption_grounding,Language-to-Time Grounding,motion_capture,Motion Capture,computed,0.030569594353437424,mrr,0.030569594353437424,,false,,,,,,,,,,,,,,,,,,0.030569594353437424,0.008620689655172414,0.02586206896551724,0.04885057471264368,110.5,130.32470703125,348
+caption_grounding,Language-to-Time Grounding,inertial,Inertial,computed,0.02470344305038452,mrr,0.02470344305038452,,false,,,,,,,,,,,,,,,,,,0.02470344305038452,0.0028735632183908046,0.022988505747126436,0.04597701149425287,123.0,138.61207580566406,348
+caption_grounding,Language-to-Time Grounding,language,Language,computed,0.24527303874492645,mrr,0.24527303874492645,,true,,,,,,,,,,,,,,,,,,0.24527303874492645,0.12643678160919541,0.34770114942528735,0.47126436781609193,12.0,15.106322288513184,348
+caption_grounding,Language-to-Time Grounding,no_language,All Except Language,computed,0.02722795307636261,mrr,0.02722795307636261,,false,,,,,,,,,,,,,,,,,,0.02722795307636261,0.005747126436781609,0.028735632183908046,0.04597701149425287,134.0,142.65516662597656,348
+cross_modal_retrieval,Cross-Modal Window Retrieval,all_features,All Features,computed,0.9723829030990601,mrr,0.9723829030990601,,true,,,,,,,,,,,,,,,,,,0.9723829030990601,0.9683908045977011,0.9741379310344828,0.9827586206896551,1.0,2.347701072692871,348
+cross_modal_retrieval,Cross-Modal Window Retrieval,video,Video,computed,0.9701701402664185,mrr,0.9701701402664185,,true,,,,,,,,,,,,,,,,,,0.9701701402664185,0.9626436781609196,0.9798850574712644,0.9798850574712644,1.0,3.844827651977539,348
+cross_modal_retrieval,Cross-Modal Window Retrieval,depth,Depth,computed,0.6656051278114319,mrr,0.6656051278114319,,true,,,,,,,,,,,,,,,,,,0.6656051278114319,0.5660919540229885,0.7902298850574713,0.8620689655172413,1.0,5.729885101318359,348
+cross_modal_retrieval,Cross-Modal Window Retrieval,pose_slam,Pose + SLAM,computed,0.42622581124305725,mrr,0.42622581124305725,,false,,,,,,,,,,,,,,,,,,0.42622581124305725,0.3017241379310345,0.5488505747126436,0.6551724137931034,4.0,15.623562812805176,348
+cross_modal_retrieval,Cross-Modal Window Retrieval,motion_capture,Motion Capture,computed,0.2553335726261139,mrr,0.2553335726261139,,false,,,,,,,,,,,,,,,,,,0.2553335726261139,0.15804597701149425,0.35344827586206895,0.3994252873563218,21.5,49.181034088134766,348
+cross_modal_retrieval,Cross-Modal Window Retrieval,inertial,Inertial,computed,0.2840072810649872,mrr,0.2840072810649872,,false,,,,,,,,,,,,,,,,,,0.2840072810649872,0.16379310344827586,0.3735632183908046,0.5229885057471264,10.0,20.577587127685547,348
+cross_modal_retrieval,Cross-Modal Window Retrieval,language,Language,computed,0.031006580218672752,mrr,0.031006580218672752,,false,,,,,,,,,,,,,,,,,,0.031006580218672752,0.005747126436781609,0.031609195402298854,0.05747126436781609,138.0,146.83045959472656,348
+cross_modal_retrieval,Cross-Modal Window Retrieval,no_language,All Except Language,computed,0.9722298979759216,mrr,0.9722298979759216,,true,,,,,,,,,,,,,,,,,,0.9722298979759216,0.9683908045977011,0.9741379310344828,0.9827586206896551,1.0,2.55747127532959,348
+modality_reconstruction,Sensor-to-Visual Reconstruction,all_features,All Features,computed,0.1979444902694729,mae,4.051921367645264,,true,,,,,,,,,,4260.24853515625,4.051921367645264,0.5054433122397289,,,,,,,,,,,,
+modality_reconstruction,Sensor-to-Visual Reconstruction,video,Video,computed,0.611318891594774,mae,0.635807454586029,,true,,,,,,,,,,8679.7548828125,0.635807454586029,-0.007601057526781085,,,,,,,,,,,,
+modality_reconstruction,Sensor-to-Visual Reconstruction,depth,Depth,computed,0.062215385980961393,mae,15.07319450378418,,true,,,,,,,,,,38000.71875,15.07319450378418,-3.4113648334674167,,,,,,,,,,,,
+modality_reconstruction,Sensor-to-Visual Reconstruction,pose_slam,Pose + SLAM,computed,0.5359235021455191,mae,0.8659379482269287,,false,,,,,,,,,,8678.9697265625,0.8659379482269287,-0.007509963078260462,,,,,,,,,,,,
+modality_reconstruction,Sensor-to-Visual Reconstruction,motion_capture,Motion Capture,computed,0.07724422027114182,mae,11.945952415466309,,false,,,,,,,,,,16462.224609375,11.945952415466309,-0.911039534454414,,,,,,,,,,,,
+modality_reconstruction,Sensor-to-Visual Reconstruction,inertial,Inertial,computed,0.5185351442505587,mae,0.9285095930099487,,false,,,,,,,,,,8680.1376953125,0.9285095930099487,-0.007645498747803181,,,,,,,,,,,,
+modality_reconstruction,Sensor-to-Visual Reconstruction,language,Language,computed,0.411308516935754,mae,1.4312649965286255,,false,,,,,,,,,,8681.390625,1.4312649965286255,-0.00779095493123938,,,,,,,,,,,,
+modality_reconstruction,Sensor-to-Visual Reconstruction,no_language,All Except Language,computed,0.19605900415414898,mae,4.100505352020264,,true,,,,,,,,,,4129.71484375,4.100505352020264,0.5205964615135674,,,,,,,,,,,,
+temporal_order,Temporal Order Verification,all_features,All Features,computed,0.4942528735632184,macro_f1,0.4942528735632184,,false,,0.4942528735632184,0.4942528735632184,0.4942528735632184,2,1624,696,,0,,,,,,,,,,,,,,,
+temporal_order,Temporal Order Verification,video,Video,computed,0.5172413793103449,macro_f1,0.5172413793103449,,false,,0.5172413793103449,0.5172413793103449,0.5172413793103449,2,1624,696,,0,,,,,,,,,,,,,,,
+temporal_order,Temporal Order Verification,depth,Depth,computed,0.49424869738982513,macro_f1,0.49424869738982513,,false,,0.4942528735632184,0.49424869738982513,0.49425287356321834,2,1624,696,,0,,,,,,,,,,,,,,,
+temporal_order,Temporal Order Verification,pose_slam,Pose + SLAM,computed,0.5258620689655172,macro_f1,0.5258620689655172,,false,,0.5258620689655172,0.5258620689655172,0.5258620689655172,2,1624,696,,0,,,,,,,,,,,,,,,
+temporal_order,Temporal Order Verification,motion_capture,Motion Capture,computed,0.4942528735632184,macro_f1,0.4942528735632184,,false,,0.4942528735632184,0.4942528735632184,0.4942528735632184,2,1624,696,,0,,,,,,,,,,,,,,,
+temporal_order,Temporal Order Verification,inertial,Inertial,computed,0.5,macro_f1,0.5,,false,,0.5,0.5,0.5,2,1624,696,,0,,,,,,,,,,,,,,,
+temporal_order,Temporal Order Verification,language,Language,computed,0.4236751152073733,macro_f1,0.4236751152073733,,false,,0.47126436781609193,0.4236751152073733,0.47126436781609193,2,1624,696,,0,,,,,,,,,,,,,,,
+temporal_order,Temporal Order Verification,no_language,All Except Language,computed,0.49856218325196366,macro_f1,0.49856218325196366,,false,,0.4985632183908046,0.49856218325196366,0.4985632183908046,2,1624,696,,0,,,,,,,,,,,,,,,
+misalignment_detection,Cross-Modal Misalignment Detection,all_features,All Features,computed,0.4134495778430919,macro_f1,0.4134495778430919,,false,,0.4436416184971098,0.4134495778430919,0.4436416184971098,2,1614,692,,0,,,,,,,,,,,,,,,
+misalignment_detection,Cross-Modal Misalignment Detection,video,Video,computed,0.49488307322727143,macro_f1,0.49488307322727143,,false,,0.4985549132947977,0.49488307322727143,0.4985549132947977,2,1614,692,,0,,,,,,,,,,,,,,,
+misalignment_detection,Cross-Modal Misalignment Detection,depth,Depth,computed,0.46659963973021656,macro_f1,0.46659963973021656,,false,,0.4682080924855491,0.46659963973021656,0.4682080924855491,2,1614,692,,0,,,,,,,,,,,,,,,
+misalignment_detection,Cross-Modal Misalignment Detection,pose_slam,Pose + SLAM,computed,0.4929686094043242,macro_f1,0.4929686094043242,,false,,0.5057803468208093,0.4929686094043242,0.5057803468208092,2,1614,692,,0,,,,,,,,,,,,,,,
+misalignment_detection,Cross-Modal Misalignment Detection,motion_capture,Motion Capture,computed,0.4133918268956141,macro_f1,0.4133918268956141,,false,,0.4638728323699422,0.4133918268956141,0.4638728323699422,2,1614,692,,0,,,,,,,,,,,,,,,
+misalignment_detection,Cross-Modal Misalignment Detection,inertial,Inertial,computed,0.48899072503396884,macro_f1,0.48899072503396884,,false,,0.49421965317919075,0.48899072503396884,0.49421965317919075,2,1614,692,,0,,,,,,,,,,,,,,,
+misalignment_detection,Cross-Modal Misalignment Detection,language,Language,computed,0.4942161609504254,macro_f1,0.4942161609504254,,false,,0.5,0.4942161609504254,0.5,2,1614,692,,0,,,,,,,,,,,,,,,
+misalignment_detection,Cross-Modal Misalignment Detection,no_language,All Except Language,computed,0.41142741665056637,macro_f1,0.41142741665056637,,false,,0.44653179190751446,0.41142741665056637,0.44653179190751446,2,1614,692,,0,,,,,,,,,,,,,,,

results/single_episode_diagnostics/modality_ablation/ablation_summary.json

@@ -0,0 +1,31 @@
+{
+  "description": "Compact ridge-head ablation over real shared_windows.npz feature blocks.",
+  "num_rows": 96,
+  "num_computed": 96,
+  "num_not_computed": 0,
+  "groups": {
+    "all_features": 8378,
+    "video": 4116,
+    "depth": 980,
+    "pose_slam": 223,
+    "motion_capture": 2121,
+    "inertial": 42,
+    "language": 896,
+    "no_language": 7482
+  },
+  "tasks": [
+    "timeline_action",
+    "timeline_subtask",
+    "transition_detection",
+    "next_action",
+    "hand_trajectory_forecast",
+    "contact_prediction",
+    "object_relevance",
+    "caption_grounding",
+    "cross_modal_retrieval",
+    "modality_reconstruction",
+    "temporal_order",
+    "misalignment_detection"
+  ],
+  "object_relevance_labels": "annotation.hdf5"
+}

results/single_episode_diagnostics/object_labels/object_vocab.json

@@ -0,0 +1,42 @@
+{
+  "vocab": [
+    "kettle",
+    "coffee dripper",
+    "scale",
+    "bottle",
+    "gooseneck kettle",
+    "digital scale",
+    "table",
+    "dripper",
+    "coffee filter",
+    "glass carafe",
+    "coffee scale",
+    "white mug",
+    "wooden scoop",
+    "coffee jar",
+    "coffee scoop",
+    "coffee container",
+    "lid",
+    "closed coffee container",
+    "water bottle",
+    "coffee mug",
+    "mug",
+    "white cup",
+    "white bottle",
+    "coffee equipment",
+    "small bottle",
+    "weighing scale",
+    "white coffee cup",
+    "digital scale with dripper",
+    "metal pitcher",
+    "carafe",
+    "milk pitcher",
+    "coffee cup",
+    "stainless steel milk pitcher",
+    "milk bottle"
+  ],
+  "num_objects": 34,
+  "source_annotation": "external_raw_sample/annotation.hdf5",
+  "source_toolkit": "HOMIE-toolkit",
+  "source_note": "Object labels were exported from a raw Xperience-10M sample annotation. The public artifact stores source type and hash instead of machine-specific file paths."
+}

results/single_episode_diagnostics/object_labels/window_object_labels.csv

@@ -0,0 +1,1162 @@
+window_index,start_frame,end_frame,center_frame,objects,object_count
+0,0,19,9,kettle|coffee dripper|scale|bottle,4
+1,5,24,14,kettle|coffee dripper|scale|bottle,4
+2,10,29,19,kettle|coffee dripper|scale|bottle,4
+3,15,34,24,kettle|coffee dripper|scale|bottle,4
+4,20,39,29,kettle|coffee dripper|scale|bottle,4
+5,25,44,34,kettle|coffee dripper|scale|bottle,4
+6,30,49,39,kettle|coffee dripper|scale|bottle,4
+7,35,54,44,kettle|coffee dripper|scale|bottle,4
+8,40,59,49,kettle|coffee dripper|scale|bottle,4
+9,45,64,54,kettle|coffee dripper|scale|bottle,4
+10,50,69,59,kettle|coffee dripper|scale|bottle,4
+11,55,74,64,kettle|coffee dripper|scale|bottle,4
+12,60,79,69,kettle|coffee dripper|scale|bottle,4
+13,65,84,74,kettle|coffee dripper|scale|bottle,4
+14,70,89,79,kettle|coffee dripper|scale|bottle,4
+15,75,94,84,kettle|coffee dripper|scale|bottle,4
+16,80,99,89,kettle|coffee dripper|scale|bottle,4
+17,85,104,94,kettle|coffee dripper|scale|bottle,4
+18,90,109,99,kettle|coffee dripper|scale|bottle,4
+19,95,114,104,kettle|coffee dripper|scale|bottle,4
+20,100,119,109,kettle|coffee dripper|scale|bottle,4
+21,105,124,114,kettle|coffee dripper|scale|bottle,4
+22,110,129,119,kettle|coffee dripper|scale|bottle,4
+23,115,134,124,kettle|coffee dripper|scale|bottle,4
+24,120,139,129,kettle,1
+25,125,144,134,kettle,1
+26,130,149,139,kettle,1
+27,135,154,144,kettle,1
+28,140,159,149,kettle,1
+29,145,164,154,kettle,1
+30,150,169,159,kettle,1
+31,155,174,164,kettle,1
+32,160,179,169,kettle,1
+33,165,184,174,kettle,1
+34,170,189,179,kettle,1
+35,175,194,184,kettle,1
+36,180,199,189,kettle,1
+37,185,204,194,kettle,1
+38,190,209,199,kettle,1
+39,195,214,204,kettle,1
+40,200,219,209,kettle,1
+41,205,224,214,kettle,1
+42,210,229,219,kettle,1
+43,215,234,224,kettle,1
+44,220,239,229,kettle,1
+45,225,244,234,kettle,1
+46,230,249,239,kettle,1
+47,235,254,244,kettle,1
+48,240,259,249,kettle,1
+49,245,264,254,kettle,1
+50,250,269,259,kettle,1
+51,255,274,264,kettle,1
+52,260,279,269,kettle,1
+53,265,284,274,kettle,1
+54,270,289,279,kettle,1
+55,275,294,284,kettle,1
+56,280,299,289,kettle,1
+57,285,304,294,kettle,1
+58,290,309,299,kettle,1
+59,295,314,304,kettle,1
+60,300,319,309,kettle,1
+61,305,324,314,kettle,1
+62,310,329,319,kettle,1
+63,315,334,324,kettle,1
+64,320,339,329,kettle,1
+65,325,344,334,kettle,1
+66,330,349,339,kettle,1
+67,335,354,344,kettle,1
+68,340,359,349,kettle,1
+69,345,364,354,kettle|coffee dripper|scale,3
+70,350,369,359,kettle|coffee dripper|scale,3
+71,355,374,364,kettle|coffee dripper|scale,3
+72,360,379,369,kettle|coffee dripper|scale,3
+73,365,384,374,kettle|coffee dripper|scale,3
+74,370,389,379,kettle|coffee dripper|scale,3
+75,375,394,384,kettle|coffee dripper|scale,3
+76,380,399,389,kettle|coffee dripper|scale,3
+77,385,404,394,kettle|coffee dripper|scale|gooseneck kettle|digital scale|table,6
+78,390,409,399,kettle|coffee dripper|scale|gooseneck kettle|digital scale|table,6
+79,395,414,404,kettle|coffee dripper|scale|gooseneck kettle|digital scale|table,6
+80,400,419,409,gooseneck kettle|coffee dripper|digital scale|table,4
+81,405,424,414,gooseneck kettle|coffee dripper|digital scale|table,4
+82,410,429,419,gooseneck kettle|coffee dripper|digital scale|table,4
+83,415,434,424,gooseneck kettle|coffee dripper|digital scale|table,4
+84,420,439,429,gooseneck kettle|coffee dripper|digital scale|table,4
+85,425,444,434,gooseneck kettle|coffee dripper|digital scale|table,4
+86,430,449,439,gooseneck kettle|coffee dripper|digital scale|table,4
+87,435,454,444,gooseneck kettle|coffee dripper|digital scale|table,4
+88,440,459,449,gooseneck kettle|coffee dripper|digital scale|table,4
+89,445,464,454,gooseneck kettle|coffee dripper|digital scale|table,4
+90,450,469,459,gooseneck kettle|coffee dripper|digital scale|table,4
+91,455,474,464,gooseneck kettle|coffee dripper|digital scale|table,4
+92,460,479,469,gooseneck kettle|coffee dripper|digital scale|table,4
+93,465,484,474,gooseneck kettle|coffee dripper|digital scale|table,4
+94,470,489,479,gooseneck kettle|coffee dripper|digital scale|table,4
+95,475,494,484,gooseneck kettle|coffee dripper|digital scale|table,4
+96,480,499,489,gooseneck kettle|coffee dripper|digital scale|table,4
+97,485,504,494,gooseneck kettle|coffee dripper|digital scale|table,4
+98,490,509,499,gooseneck kettle|coffee dripper|digital scale|table,4
+99,495,514,504,gooseneck kettle|coffee dripper|digital scale|table,4
+100,500,519,509,gooseneck kettle|coffee dripper|digital scale|table,4
+101,505,524,514,gooseneck kettle|coffee dripper|digital scale|table,4
+102,510,529,519,gooseneck kettle|coffee dripper|digital scale|table,4
+103,515,534,524,gooseneck kettle|coffee dripper|digital scale|table,4
+104,520,539,529,gooseneck kettle|coffee dripper|digital scale,3
+105,525,544,534,gooseneck kettle|coffee dripper|digital scale,3
+106,530,549,539,gooseneck kettle|coffee dripper|digital scale,3
+107,535,554,544,gooseneck kettle|coffee dripper|digital scale,3
+108,540,559,549,gooseneck kettle|coffee dripper|digital scale,3
+109,545,564,554,gooseneck kettle|coffee dripper|digital scale,3
+110,550,569,559,gooseneck kettle|coffee dripper|digital scale,3
+111,555,574,564,gooseneck kettle|coffee dripper|digital scale,3
+112,560,579,569,gooseneck kettle|coffee dripper|digital scale,3
+113,565,584,574,gooseneck kettle|coffee dripper|digital scale,3
+114,570,589,579,gooseneck kettle|coffee dripper|digital scale,3
+115,575,594,584,gooseneck kettle|coffee dripper|digital scale,3
+116,580,599,589,gooseneck kettle|coffee dripper|digital scale,3
+117,585,604,594,gooseneck kettle|coffee dripper|digital scale,3
+118,590,609,599,gooseneck kettle|coffee dripper|digital scale,3
+119,595,614,604,gooseneck kettle|coffee dripper|digital scale,3
+120,600,619,609,gooseneck kettle|coffee dripper|digital scale,3
+121,605,624,614,gooseneck kettle|coffee dripper|digital scale,3
+122,610,629,619,gooseneck kettle|coffee dripper|digital scale,3
+123,615,634,624,gooseneck kettle|coffee dripper|digital scale,3
+124,620,639,629,gooseneck kettle|coffee dripper|digital scale,3
+125,625,644,634,gooseneck kettle|coffee dripper|digital scale,3
+126,630,649,639,gooseneck kettle|coffee dripper|digital scale,3
+127,635,654,644,gooseneck kettle|coffee dripper|digital scale,3
+128,640,659,649,gooseneck kettle|coffee dripper|digital scale,3
+129,645,664,654,gooseneck kettle|coffee dripper|digital scale,3
+130,650,669,659,gooseneck kettle|coffee dripper|digital scale,3
+131,655,674,664,gooseneck kettle|coffee dripper|digital scale,3
+132,660,679,669,gooseneck kettle|coffee dripper|digital scale,3
+133,665,684,674,gooseneck kettle|coffee dripper|digital scale,3
+134,670,689,679,gooseneck kettle|coffee dripper|digital scale,3
+135,675,694,684,gooseneck kettle|coffee dripper|digital scale,3
+136,680,699,689,gooseneck kettle|coffee dripper|digital scale,3
+137,685,704,694,gooseneck kettle|coffee dripper|digital scale,3
+138,690,709,699,gooseneck kettle|coffee dripper|digital scale,3
+139,695,714,704,gooseneck kettle|coffee dripper|digital scale,3
+140,700,719,709,gooseneck kettle|coffee dripper|digital scale,3
+141,705,724,714,gooseneck kettle|coffee dripper|digital scale,3
+142,710,729,719,gooseneck kettle|coffee dripper|digital scale,3
+143,715,734,724,gooseneck kettle|coffee dripper|digital scale,3
+144,720,739,729,gooseneck kettle|coffee dripper|digital scale,3
+145,725,744,734,gooseneck kettle|coffee dripper|digital scale,3
+146,730,749,739,gooseneck kettle|coffee dripper|digital scale,3
+147,735,754,744,gooseneck kettle|coffee dripper|digital scale,3
+148,740,759,749,gooseneck kettle|coffee dripper|digital scale,3
+149,745,764,754,gooseneck kettle|coffee dripper|digital scale,3
+150,750,769,759,gooseneck kettle|coffee dripper|digital scale,3
+151,755,774,764,gooseneck kettle|coffee dripper|digital scale,3
+152,760,779,769,gooseneck kettle|coffee dripper|digital scale,3
+153,765,784,774,gooseneck kettle|coffee dripper|digital scale,3
+154,770,789,779,gooseneck kettle|coffee dripper|digital scale,3
+155,775,794,784,gooseneck kettle|coffee dripper|digital scale,3
+156,780,799,789,gooseneck kettle|coffee dripper|digital scale,3
+157,785,804,794,gooseneck kettle|coffee dripper|digital scale|kettle|dripper|scale|coffee filter|table,8
+158,790,809,799,gooseneck kettle|coffee dripper|digital scale|kettle|dripper|scale|coffee filter|table,8
+159,795,814,804,gooseneck kettle|coffee dripper|digital scale|kettle|dripper|scale|coffee filter|table,8
+160,800,819,809,kettle|dripper|scale|coffee filter|table,5
+161,805,824,814,kettle|dripper|scale|coffee filter|table,5
+162,810,829,819,kettle|dripper|scale|coffee filter|table,5
+163,815,834,824,kettle|dripper|scale|coffee filter|table,5
+164,820,839,829,kettle|dripper|scale|coffee filter|table,5
+165,825,844,834,kettle|dripper|scale|coffee filter|table,5
+166,830,849,839,kettle|dripper|scale|coffee filter|table,5
+167,835,854,844,kettle|dripper|scale|coffee filter|table,5
+168,840,859,849,kettle|dripper|scale|coffee filter|table,5
+169,845,864,854,kettle|dripper|scale|coffee filter|table,5
+170,850,869,859,kettle|dripper|scale|coffee filter|table,5
+171,855,874,864,kettle|dripper|scale|coffee filter|table,5
+172,860,879,869,kettle|dripper|scale|coffee filter|table,5
+173,865,884,874,kettle|dripper|scale|coffee filter|table,5
+174,870,889,879,kettle|dripper|scale|coffee filter|table,5
+175,875,894,884,kettle|dripper|scale|coffee filter|table,5
+176,880,899,889,kettle|dripper|scale|coffee filter|table,5
+177,885,904,894,kettle|dripper|scale|coffee filter|table,5
+178,890,909,899,kettle|dripper|scale|coffee filter|table,5
+179,895,914,904,kettle|dripper|scale|coffee filter|table,5
+180,900,919,909,kettle|dripper|scale|coffee filter|table,5
+181,905,924,914,kettle|dripper|scale|coffee filter|table,5
+182,910,929,919,kettle|dripper|scale|coffee filter|table,5
+183,915,934,924,kettle|dripper|scale|coffee filter|table,5
+184,920,939,929,kettle|dripper|scale|coffee filter|table,5
+185,925,944,934,kettle|dripper|scale|coffee filter|table,5
+186,930,949,939,kettle|dripper|scale|coffee filter|table,5
+187,935,954,944,kettle|dripper|scale|coffee filter|table,5
+188,940,959,949,kettle|dripper|scale|coffee filter|table,5
+189,945,964,954,kettle|dripper|scale|coffee filter|table,5
+190,950,969,959,kettle|dripper|scale|coffee filter|table,5
+191,955,974,964,kettle|dripper|scale|coffee filter|table,5
+192,960,979,969,kettle|dripper|scale|coffee filter|table,5
+193,965,984,974,kettle|dripper|scale|coffee filter|table,5
+194,970,989,979,kettle|dripper|scale|coffee filter|table,5
+195,975,994,984,kettle|dripper|scale|coffee filter|table,5
+196,980,999,989,kettle|dripper|scale|coffee filter|table,5
+197,985,1004,994,kettle|dripper|scale|coffee filter|table|glass carafe|coffee scale|coffee dripper|white mug|bottle,10
+198,990,1009,999,kettle|dripper|scale|coffee filter|table|glass carafe|coffee scale|coffee dripper|white mug|bottle,10
+199,995,1014,1004,kettle|dripper|scale|coffee filter|table|glass carafe|coffee scale|coffee dripper|white mug|bottle,10
+200,1000,1019,1009,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+201,1005,1024,1014,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+202,1010,1029,1019,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+203,1015,1034,1024,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+204,1020,1039,1029,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+205,1025,1044,1034,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+206,1030,1049,1039,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+207,1035,1054,1044,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+208,1040,1059,1049,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+209,1045,1064,1054,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+210,1050,1069,1059,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+211,1055,1074,1064,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+212,1060,1079,1069,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+213,1065,1084,1074,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+214,1070,1089,1079,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+215,1075,1094,1084,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+216,1080,1099,1089,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+217,1085,1104,1094,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+218,1090,1109,1099,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+219,1095,1114,1104,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+220,1100,1119,1109,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+221,1105,1124,1114,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+222,1110,1129,1119,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+223,1115,1134,1124,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+224,1120,1139,1129,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+225,1125,1144,1134,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+226,1130,1149,1139,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+227,1135,1154,1144,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+228,1140,1159,1149,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+229,1145,1164,1154,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+230,1150,1169,1159,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+231,1155,1174,1164,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+232,1160,1179,1169,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+233,1165,1184,1174,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+234,1170,1189,1179,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+235,1175,1194,1184,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+236,1180,1199,1189,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+237,1185,1204,1194,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+238,1190,1209,1199,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+239,1195,1214,1204,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+240,1200,1219,1209,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+241,1205,1224,1214,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+242,1210,1229,1219,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+243,1215,1234,1224,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+244,1220,1239,1229,glass carafe|coffee scale|coffee dripper|white mug|bottle,5
+245,1225,1244,1234,glass carafe|coffee scale|coffee dripper|white mug|bottle|wooden scoop,6
+246,1230,1249,1239,glass carafe|coffee scale|coffee dripper|white mug|bottle|wooden scoop,6
+247,1235,1254,1244,glass carafe|coffee scale|coffee dripper|white mug|bottle|wooden scoop,6
+248,1240,1259,1249,glass carafe|wooden scoop,2
+249,1245,1264,1254,glass carafe|wooden scoop,2
+250,1250,1269,1259,glass carafe|wooden scoop,2
+251,1255,1274,1264,glass carafe|wooden scoop,2
+252,1260,1279,1269,glass carafe|wooden scoop,2
+253,1265,1284,1274,glass carafe|wooden scoop,2
+254,1270,1289,1279,glass carafe|wooden scoop,2
+255,1275,1294,1284,glass carafe|wooden scoop,2
+256,1280,1299,1289,glass carafe|wooden scoop,2
+257,1285,1304,1294,glass carafe|wooden scoop,2
+258,1290,1309,1299,glass carafe|wooden scoop,2
+259,1295,1314,1304,glass carafe|wooden scoop,2
+260,1300,1319,1309,glass carafe|wooden scoop,2
+261,1305,1324,1314,glass carafe|wooden scoop,2
+262,1310,1329,1319,glass carafe|wooden scoop,2
+263,1315,1334,1324,glass carafe|wooden scoop,2
+264,1320,1339,1329,glass carafe|wooden scoop,2
+265,1325,1344,1334,glass carafe|wooden scoop,2
+266,1330,1349,1339,glass carafe|wooden scoop,2
+267,1335,1354,1344,glass carafe|wooden scoop,2
+268,1340,1359,1349,glass carafe|wooden scoop,2
+269,1345,1364,1354,glass carafe|wooden scoop,2
+270,1350,1369,1359,glass carafe|wooden scoop,2
+271,1355,1374,1364,glass carafe|wooden scoop,2
+272,1360,1379,1369,glass carafe|wooden scoop,2
+273,1365,1384,1374,glass carafe|wooden scoop,2
+274,1370,1389,1379,glass carafe|wooden scoop,2
+275,1375,1394,1384,glass carafe|wooden scoop,2
+276,1380,1399,1389,glass carafe|wooden scoop,2
+277,1385,1404,1394,glass carafe|wooden scoop|coffee jar|coffee scoop|dripper,5
+278,1390,1409,1399,glass carafe|wooden scoop|coffee jar|coffee scoop|dripper,5
+279,1395,1414,1404,glass carafe|wooden scoop|coffee jar|coffee scoop|dripper,5
+280,1400,1419,1409,coffee jar|coffee scoop|dripper,3
+281,1405,1424,1414,coffee jar|coffee scoop|dripper,3
+282,1410,1429,1419,coffee jar|coffee scoop|dripper,3
+283,1415,1434,1424,coffee jar|coffee scoop|dripper,3
+284,1420,1439,1429,coffee jar|coffee scoop|dripper,3
+285,1425,1444,1434,coffee jar|coffee scoop|dripper,3
+286,1430,1449,1439,coffee jar|coffee scoop|dripper,3
+287,1435,1454,1444,coffee jar|coffee scoop|dripper,3
+288,1440,1459,1449,coffee jar|coffee scoop|dripper,3
+289,1445,1464,1454,coffee jar|coffee scoop|dripper,3
+290,1450,1469,1459,coffee jar|coffee scoop|dripper,3
+291,1455,1474,1464,coffee jar|coffee scoop|dripper,3
+292,1460,1479,1469,coffee jar|coffee scoop|dripper,3
+293,1465,1484,1474,coffee jar|coffee scoop|dripper,3
+294,1470,1489,1479,coffee jar|coffee scoop|dripper,3
+295,1475,1494,1484,coffee jar|coffee scoop|dripper,3
+296,1480,1499,1489,coffee jar|coffee scoop,2
+297,1485,1504,1494,coffee jar|coffee scoop,2
+298,1490,1509,1499,coffee jar|coffee scoop,2
+299,1495,1514,1504,coffee jar|coffee scoop,2
+300,1500,1519,1509,coffee jar|coffee scoop,2
+301,1505,1524,1514,coffee jar|coffee scoop,2
+302,1510,1529,1519,coffee jar|coffee scoop,2
+303,1515,1534,1524,coffee jar|coffee scoop,2
+304,1520,1539,1529,coffee jar|coffee scoop,2
+305,1525,1544,1534,coffee jar|coffee scoop,2
+306,1530,1549,1539,coffee jar|coffee scoop,2
+307,1535,1554,1544,coffee jar|coffee scoop,2
+308,1540,1559,1549,coffee jar|coffee scoop,2
+309,1545,1564,1554,coffee jar|coffee scoop,2
+310,1550,1569,1559,coffee jar|coffee scoop,2
+311,1555,1574,1564,coffee jar|coffee scoop,2
+312,1560,1579,1569,coffee jar|coffee scoop,2
+313,1565,1584,1574,coffee jar|coffee scoop,2
+314,1570,1589,1579,coffee jar|coffee scoop,2
+315,1575,1594,1584,coffee jar|coffee scoop,2
+316,1580,1599,1589,coffee jar|coffee scoop,2
+317,1585,1604,1594,coffee jar|coffee scoop|dripper,3
+318,1590,1609,1599,coffee jar|coffee scoop|dripper,3
+319,1595,1614,1604,coffee jar|coffee scoop|dripper,3
+320,1600,1619,1609,coffee scoop|dripper,2
+321,1605,1624,1614,coffee scoop|dripper,2
+322,1610,1629,1619,coffee scoop|dripper,2
+323,1615,1634,1624,coffee scoop|dripper,2
+324,1620,1639,1629,coffee scoop|dripper,2
+325,1625,1644,1634,coffee scoop|dripper,2
+326,1630,1649,1639,coffee scoop|dripper,2
+327,1635,1654,1644,coffee scoop|dripper,2
+328,1640,1659,1649,coffee scoop|dripper,2
+329,1645,1664,1654,coffee scoop|dripper,2
+330,1650,1669,1659,coffee scoop|dripper,2
+331,1655,1674,1664,coffee scoop|dripper,2
+332,1660,1679,1669,coffee scoop|dripper,2
+333,1665,1684,1674,coffee scoop|dripper,2
+334,1670,1689,1679,coffee scoop|dripper,2
+335,1675,1694,1684,coffee scoop|dripper,2
+336,1680,1699,1689,coffee scoop|dripper,2
+337,1685,1704,1694,coffee scoop|dripper,2
+338,1690,1709,1699,coffee scoop|dripper,2
+339,1695,1714,1704,coffee scoop|dripper,2
+340,1700,1719,1709,coffee scoop|dripper,2
+341,1705,1724,1714,coffee scoop|dripper,2
+342,1710,1729,1719,coffee scoop|dripper,2
+343,1715,1734,1724,coffee scoop|dripper,2
+344,1720,1739,1729,coffee scoop|dripper,2
+345,1725,1744,1734,coffee scoop|dripper,2
+346,1730,1749,1739,coffee scoop|dripper,2
+347,1735,1754,1744,coffee scoop|dripper,2
+348,1740,1759,1749,coffee scoop|dripper,2
+349,1745,1764,1754,coffee scoop|dripper,2
+350,1750,1769,1759,coffee scoop|dripper,2
+351,1755,1774,1764,coffee scoop|dripper,2
+352,1760,1779,1769,coffee scoop|dripper,2
+353,1765,1784,1774,coffee scoop|dripper,2
+354,1770,1789,1779,coffee scoop|dripper,2
+355,1775,1794,1784,coffee scoop|dripper,2
+356,1780,1799,1789,coffee scoop|dripper,2
+357,1785,1804,1794,coffee scoop|dripper|coffee container|lid|scale,5
+358,1790,1809,1799,coffee scoop|dripper|coffee container|lid|scale,5
+359,1795,1814,1804,coffee scoop|dripper|coffee container|lid|scale,5
+360,1800,1819,1809,coffee container|lid|coffee scoop|dripper|scale,5
+361,1805,1824,1814,coffee container|lid|coffee scoop|dripper|scale,5
+362,1810,1829,1819,coffee container|lid|coffee scoop|dripper|scale,5
+363,1815,1834,1824,coffee container|lid|coffee scoop|dripper|scale,5
+364,1820,1839,1829,coffee container|lid|coffee scoop|dripper|scale,5
+365,1825,1844,1834,coffee container|lid|coffee scoop|dripper|scale,5
+366,1830,1849,1839,coffee container|lid|coffee scoop|dripper|scale,5
+367,1835,1854,1844,coffee container|lid|coffee scoop|dripper|scale,5
+368,1840,1859,1849,coffee container|lid|coffee scoop|dripper|scale,5
+369,1845,1864,1854,coffee container|lid|coffee scoop|dripper|scale,5
+370,1850,1869,1859,coffee container|lid|coffee scoop|dripper|scale,5
+371,1855,1874,1864,coffee container|lid|coffee scoop|dripper|scale,5
+372,1860,1879,1869,coffee container|lid|coffee scoop|dripper|scale,5
+373,1865,1884,1874,coffee container|lid|coffee scoop|dripper|scale,5
+374,1870,1889,1879,coffee container|lid|coffee scoop|dripper|scale,5
+375,1875,1894,1884,coffee container|lid|coffee scoop|dripper|scale,5
+376,1880,1899,1889,coffee container|lid|coffee scoop|dripper|scale,5
+377,1885,1904,1894,coffee container|lid|coffee scoop|dripper|scale,5
+378,1890,1909,1899,coffee container|lid|coffee scoop|dripper|scale,5
+379,1895,1914,1904,coffee container|lid|coffee scoop|dripper|scale,5
+380,1900,1919,1909,coffee container|lid|coffee scoop|dripper|scale,5
+381,1905,1924,1914,coffee container|lid|coffee scoop|dripper|scale,5
+382,1910,1929,1919,coffee container|lid|coffee scoop|dripper|scale,5
+383,1915,1934,1924,coffee container|lid|coffee scoop|dripper|scale,5
+384,1920,1939,1929,coffee container|lid|coffee scoop|dripper|scale,5
+385,1925,1944,1934,coffee container|lid|coffee scoop|dripper|scale,5
+386,1930,1949,1939,coffee container|lid|coffee scoop|dripper|scale,5
+387,1935,1954,1944,coffee container|lid|coffee scoop|dripper|scale,5
+388,1940,1959,1949,coffee container|lid|coffee scoop|dripper|scale,5
+389,1945,1964,1954,coffee container|lid|coffee scoop|dripper|scale|closed coffee container,6
+390,1950,1969,1959,coffee container|lid|coffee scoop|dripper|scale|closed coffee container,6
+391,1955,1974,1964,coffee container|lid|coffee scoop|dripper|scale|closed coffee container,6
+392,1960,1979,1969,closed coffee container|dripper|scale,3
+393,1965,1984,1974,closed coffee container|dripper|scale,3
+394,1970,1989,1979,closed coffee container|dripper|scale,3
+395,1975,1994,1984,closed coffee container|dripper|scale,3
+396,1980,1999,1989,closed coffee container|dripper|scale,3
+397,1985,2004,1994,closed coffee container|dripper|scale|gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,8
+398,1990,2009,1999,closed coffee container|dripper|scale|gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,8
+399,1995,2014,2004,closed coffee container|dripper|scale|gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,8
+400,2000,2019,2009,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+401,2005,2024,2014,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+402,2010,2029,2019,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+403,2015,2034,2024,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+404,2020,2039,2029,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+405,2025,2044,2034,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+406,2030,2049,2039,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+407,2035,2054,2044,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+408,2040,2059,2049,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+409,2045,2064,2054,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+410,2050,2069,2059,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+411,2055,2074,2064,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+412,2060,2079,2069,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+413,2065,2084,2074,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+414,2070,2089,2079,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+415,2075,2094,2084,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+416,2080,2099,2089,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+417,2085,2104,2094,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+418,2090,2109,2099,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+419,2095,2114,2104,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+420,2100,2119,2109,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+421,2105,2124,2114,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+422,2110,2129,2119,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+423,2115,2134,2124,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+424,2120,2139,2129,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+425,2125,2144,2134,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+426,2130,2149,2139,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+427,2135,2154,2144,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+428,2140,2159,2149,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+429,2145,2164,2154,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+430,2150,2169,2159,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+431,2155,2174,2164,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+432,2160,2179,2169,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+433,2165,2184,2174,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+434,2170,2189,2179,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+435,2175,2194,2184,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+436,2180,2199,2189,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+437,2185,2204,2194,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+438,2190,2209,2199,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+439,2195,2214,2204,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+440,2200,2219,2209,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+441,2205,2224,2214,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+442,2210,2229,2219,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+443,2215,2234,2224,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+444,2220,2239,2229,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+445,2225,2244,2234,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+446,2230,2249,2239,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+447,2235,2254,2244,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+448,2240,2259,2249,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+449,2245,2264,2254,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+450,2250,2269,2259,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+451,2255,2274,2264,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+452,2260,2279,2269,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+453,2265,2284,2274,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+454,2270,2289,2279,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+455,2275,2294,2284,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+456,2280,2299,2289,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+457,2285,2304,2294,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+458,2290,2309,2299,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+459,2295,2314,2304,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+460,2300,2319,2309,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+461,2305,2324,2314,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+462,2310,2329,2319,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+463,2315,2334,2324,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+464,2320,2339,2329,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+465,2325,2344,2334,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+466,2330,2349,2339,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+467,2335,2354,2344,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+468,2340,2359,2349,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+469,2345,2364,2354,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+470,2350,2369,2359,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+471,2355,2374,2364,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+472,2360,2379,2369,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+473,2365,2384,2374,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+474,2370,2389,2379,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+475,2375,2394,2384,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+476,2380,2399,2389,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug,5
+477,2385,2404,2394,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug|scale|mug,7
+478,2390,2409,2399,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug|scale|mug,7
+479,2395,2414,2404,gooseneck kettle|coffee dripper|digital scale|water bottle|coffee mug|scale|mug,7
+480,2400,2419,2409,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+481,2405,2424,2414,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+482,2410,2429,2419,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+483,2415,2434,2424,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+484,2420,2439,2429,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+485,2425,2444,2434,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+486,2430,2449,2439,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+487,2435,2454,2444,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+488,2440,2459,2449,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+489,2445,2464,2454,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+490,2450,2469,2459,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+491,2455,2474,2464,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+492,2460,2479,2469,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+493,2465,2484,2474,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+494,2470,2489,2479,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+495,2475,2494,2484,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+496,2480,2499,2489,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+497,2485,2504,2494,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+498,2490,2509,2499,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+499,2495,2514,2504,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+500,2500,2519,2509,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+501,2505,2524,2514,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+502,2510,2529,2519,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+503,2515,2534,2524,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+504,2520,2539,2529,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+505,2525,2544,2534,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+506,2530,2549,2539,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+507,2535,2554,2544,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+508,2540,2559,2549,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+509,2545,2564,2554,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+510,2550,2569,2559,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+511,2555,2574,2564,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+512,2560,2579,2569,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+513,2565,2584,2574,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+514,2570,2589,2579,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+515,2575,2594,2584,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+516,2580,2599,2589,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+517,2585,2604,2594,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+518,2590,2609,2599,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+519,2595,2614,2604,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+520,2600,2619,2609,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+521,2605,2624,2614,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+522,2610,2629,2619,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+523,2615,2634,2624,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+524,2620,2639,2629,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+525,2625,2644,2634,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+526,2630,2649,2639,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+527,2635,2654,2644,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+528,2640,2659,2649,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+529,2645,2664,2654,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+530,2650,2669,2659,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+531,2655,2674,2664,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+532,2660,2679,2669,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+533,2665,2684,2674,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+534,2670,2689,2679,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+535,2675,2694,2684,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+536,2680,2699,2689,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+537,2685,2704,2694,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+538,2690,2709,2699,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+539,2695,2714,2704,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+540,2700,2719,2709,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+541,2705,2724,2714,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+542,2710,2729,2719,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+543,2715,2734,2724,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+544,2720,2739,2729,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+545,2725,2744,2734,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+546,2730,2749,2739,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+547,2735,2754,2744,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+548,2740,2759,2749,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+549,2745,2764,2754,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+550,2750,2769,2759,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+551,2755,2774,2764,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+552,2760,2779,2769,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+553,2765,2784,2774,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+554,2770,2789,2779,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+555,2775,2794,2784,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+556,2780,2799,2789,gooseneck kettle|coffee dripper|scale|mug|water bottle,5
+557,2785,2804,2794,gooseneck kettle|coffee dripper|scale|mug|water bottle|white cup,6
+558,2790,2809,2799,gooseneck kettle|coffee dripper|scale|mug|water bottle|white cup,6
+559,2795,2814,2804,gooseneck kettle|coffee dripper|scale|mug|water bottle|white cup,6
+560,2800,2819,2809,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+561,2805,2824,2814,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+562,2810,2829,2819,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+563,2815,2834,2824,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+564,2820,2839,2829,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+565,2825,2844,2834,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+566,2830,2849,2839,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+567,2835,2854,2844,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+568,2840,2859,2849,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+569,2845,2864,2854,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+570,2850,2869,2859,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+571,2855,2874,2864,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+572,2860,2879,2869,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+573,2865,2884,2874,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+574,2870,2889,2879,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+575,2875,2894,2884,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+576,2880,2899,2889,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+577,2885,2904,2894,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+578,2890,2909,2899,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+579,2895,2914,2904,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+580,2900,2919,2909,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+581,2905,2924,2914,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+582,2910,2929,2919,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+583,2915,2934,2924,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+584,2920,2939,2929,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+585,2925,2944,2934,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+586,2930,2949,2939,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+587,2935,2954,2944,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+588,2940,2959,2949,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+589,2945,2964,2954,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+590,2950,2969,2959,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+591,2955,2974,2964,gooseneck kettle|coffee dripper|scale|water bottle|white cup,5
+592,2960,2979,2969,gooseneck kettle|coffee dripper|scale,3
+593,2965,2984,2974,gooseneck kettle|coffee dripper|scale,3
+594,2970,2989,2979,gooseneck kettle|coffee dripper|scale,3
+595,2975,2994,2984,gooseneck kettle|coffee dripper|scale,3
+596,2980,2999,2989,gooseneck kettle|coffee dripper|scale,3
+597,2985,3004,2994,gooseneck kettle|coffee dripper|scale|digital scale|white mug,5
+598,2990,3009,2999,gooseneck kettle|coffee dripper|scale|digital scale|white mug,5
+599,2995,3014,3004,gooseneck kettle|coffee dripper|scale|digital scale|white mug,5
+600,3000,3019,3009,gooseneck kettle|coffee dripper|digital scale|white mug,4
+601,3005,3024,3014,gooseneck kettle|coffee dripper|digital scale|white mug,4
+602,3010,3029,3019,gooseneck kettle|coffee dripper|digital scale|white mug,4
+603,3015,3034,3024,gooseneck kettle|coffee dripper|digital scale|white mug,4
+604,3020,3039,3029,gooseneck kettle|coffee dripper|digital scale|white mug,4
+605,3025,3044,3034,gooseneck kettle|coffee dripper|digital scale|white mug,4
+606,3030,3049,3039,gooseneck kettle|coffee dripper|digital scale|white mug,4
+607,3035,3054,3044,gooseneck kettle|coffee dripper|digital scale|white mug,4
+608,3040,3059,3049,gooseneck kettle|coffee dripper|digital scale|white mug,4
+609,3045,3064,3054,gooseneck kettle|coffee dripper|digital scale|white mug,4
+610,3050,3069,3059,gooseneck kettle|coffee dripper|digital scale|white mug,4
+611,3055,3074,3064,gooseneck kettle|coffee dripper|digital scale|white mug,4
+612,3060,3079,3069,gooseneck kettle|coffee dripper|digital scale|white mug,4
+613,3065,3084,3074,gooseneck kettle|coffee dripper|digital scale|white mug,4
+614,3070,3089,3079,gooseneck kettle|coffee dripper|digital scale|white mug,4
+615,3075,3094,3084,gooseneck kettle|coffee dripper|digital scale|white mug,4
+616,3080,3099,3089,gooseneck kettle|coffee dripper|digital scale|white mug,4
+617,3085,3104,3094,gooseneck kettle|coffee dripper|digital scale|white mug,4
+618,3090,3109,3099,gooseneck kettle|coffee dripper|digital scale|white mug,4
+619,3095,3114,3104,gooseneck kettle|coffee dripper|digital scale|white mug,4
+620,3100,3119,3109,gooseneck kettle|coffee dripper|digital scale|white mug,4
+621,3105,3124,3114,gooseneck kettle|coffee dripper|digital scale|white mug,4
+622,3110,3129,3119,gooseneck kettle|coffee dripper|digital scale|white mug,4
+623,3115,3134,3124,gooseneck kettle|coffee dripper|digital scale|white mug,4
+624,3120,3139,3129,gooseneck kettle|coffee dripper|digital scale|white mug,4
+625,3125,3144,3134,gooseneck kettle|coffee dripper|digital scale|white mug,4
+626,3130,3149,3139,gooseneck kettle|coffee dripper|digital scale|white mug,4
+627,3135,3154,3144,gooseneck kettle|coffee dripper|digital scale|white mug,4
+628,3140,3159,3149,gooseneck kettle|coffee dripper|digital scale|white mug,4
+629,3145,3164,3154,gooseneck kettle|coffee dripper|digital scale|white mug,4
+630,3150,3169,3159,gooseneck kettle|coffee dripper|digital scale|white mug,4
+631,3155,3174,3164,gooseneck kettle|coffee dripper|digital scale|white mug,4
+632,3160,3179,3169,gooseneck kettle|coffee dripper|digital scale|white mug,4
+633,3165,3184,3174,gooseneck kettle|coffee dripper|digital scale|white mug,4
+634,3170,3189,3179,gooseneck kettle|coffee dripper|digital scale|white mug,4
+635,3175,3194,3184,gooseneck kettle|coffee dripper|digital scale|white mug,4
+636,3180,3199,3189,gooseneck kettle|coffee dripper|digital scale|white mug,4
+637,3185,3204,3194,gooseneck kettle|coffee dripper|digital scale|white mug,4
+638,3190,3209,3199,gooseneck kettle|coffee dripper|digital scale|white mug,4
+639,3195,3214,3204,gooseneck kettle|coffee dripper|digital scale|white mug,4
+640,3200,3219,3209,gooseneck kettle|coffee dripper|digital scale|white mug,4
+641,3205,3224,3214,gooseneck kettle|coffee dripper|digital scale|white mug,4
+642,3210,3229,3219,gooseneck kettle|coffee dripper|digital scale|white mug,4
+643,3215,3234,3224,gooseneck kettle|coffee dripper|digital scale|white mug,4
+644,3220,3239,3229,gooseneck kettle|coffee dripper|digital scale|white mug,4
+645,3225,3244,3234,gooseneck kettle|coffee dripper|digital scale|white mug,4
+646,3230,3249,3239,gooseneck kettle|coffee dripper|digital scale|white mug,4
+647,3235,3254,3244,gooseneck kettle|coffee dripper|digital scale|white mug,4
+648,3240,3259,3249,gooseneck kettle|coffee dripper|digital scale|white mug,4
+649,3245,3264,3254,gooseneck kettle|coffee dripper|digital scale|white mug,4
+650,3250,3269,3259,gooseneck kettle|coffee dripper|digital scale|white mug,4
+651,3255,3274,3264,gooseneck kettle|coffee dripper|digital scale|white mug,4
+652,3260,3279,3269,gooseneck kettle|coffee dripper|digital scale|white mug,4
+653,3265,3284,3274,gooseneck kettle|coffee dripper|digital scale|white mug,4
+654,3270,3289,3279,gooseneck kettle|coffee dripper|digital scale|white mug,4
+655,3275,3294,3284,gooseneck kettle|coffee dripper|digital scale|white mug,4
+656,3280,3299,3289,gooseneck kettle|coffee dripper|digital scale|white mug,4
+657,3285,3304,3294,gooseneck kettle|coffee dripper|digital scale|white mug,4
+658,3290,3309,3299,gooseneck kettle|coffee dripper|digital scale|white mug,4
+659,3295,3314,3304,gooseneck kettle|coffee dripper|digital scale|white mug,4
+660,3300,3319,3309,gooseneck kettle|coffee dripper|digital scale|white mug,4
+661,3305,3324,3314,gooseneck kettle|coffee dripper|digital scale|white mug,4
+662,3310,3329,3319,gooseneck kettle|coffee dripper|digital scale|white mug,4
+663,3315,3334,3324,gooseneck kettle|coffee dripper|digital scale|white mug,4
+664,3320,3339,3329,gooseneck kettle|coffee dripper|digital scale|white mug,4
+665,3325,3344,3334,gooseneck kettle|coffee dripper|digital scale|white mug,4
+666,3330,3349,3339,gooseneck kettle|coffee dripper|digital scale|white mug,4
+667,3335,3354,3344,gooseneck kettle|coffee dripper|digital scale|white mug,4
+668,3340,3359,3349,gooseneck kettle|coffee dripper|digital scale|white mug,4
+669,3345,3364,3354,gooseneck kettle|coffee dripper|digital scale|white mug,4
+670,3350,3369,3359,gooseneck kettle|coffee dripper|digital scale|white mug,4
+671,3355,3374,3364,gooseneck kettle|coffee dripper|digital scale|white mug,4
+672,3360,3379,3369,gooseneck kettle|coffee dripper|digital scale|white mug,4
+673,3365,3384,3374,gooseneck kettle|coffee dripper|digital scale|white mug,4
+674,3370,3389,3379,gooseneck kettle|coffee dripper|digital scale|white mug,4
+675,3375,3394,3384,gooseneck kettle|coffee dripper|digital scale|white mug,4
+676,3380,3399,3389,gooseneck kettle|coffee dripper|digital scale|white mug,4
+677,3385,3404,3394,gooseneck kettle|coffee dripper|digital scale|white mug|scale|white bottle,6
+678,3390,3409,3399,gooseneck kettle|coffee dripper|digital scale|white mug|scale|white bottle,6
+679,3395,3414,3404,gooseneck kettle|coffee dripper|digital scale|white mug|scale|white bottle,6
+680,3400,3419,3409,gooseneck kettle|coffee dripper|scale|white bottle,4
+681,3405,3424,3414,gooseneck kettle|coffee dripper|scale|white bottle,4
+682,3410,3429,3419,gooseneck kettle|coffee dripper|scale|white bottle,4
+683,3415,3434,3424,gooseneck kettle|coffee dripper|scale|white bottle,4
+684,3420,3439,3429,gooseneck kettle|coffee dripper|scale|white bottle,4
+685,3425,3444,3434,gooseneck kettle|coffee dripper|scale|white bottle,4
+686,3430,3449,3439,gooseneck kettle|coffee dripper|scale|white bottle,4
+687,3435,3454,3444,gooseneck kettle|coffee dripper|scale|white bottle,4
+688,3440,3459,3449,gooseneck kettle|coffee dripper|scale|white bottle,4
+689,3445,3464,3454,gooseneck kettle|coffee dripper|scale|white bottle,4
+690,3450,3469,3459,gooseneck kettle|coffee dripper|scale|white bottle,4
+691,3455,3474,3464,gooseneck kettle|coffee dripper|scale|white bottle,4
+692,3460,3479,3469,gooseneck kettle|coffee dripper|scale|white bottle,4
+693,3465,3484,3474,gooseneck kettle|coffee dripper|scale|white bottle,4
+694,3470,3489,3479,gooseneck kettle|coffee dripper|scale|white bottle,4
+695,3475,3494,3484,gooseneck kettle|coffee dripper|scale|white bottle,4
+696,3480,3499,3489,gooseneck kettle|coffee dripper|scale|white bottle,4
+697,3485,3504,3494,gooseneck kettle|coffee dripper|scale|white bottle,4
+698,3490,3509,3499,gooseneck kettle|coffee dripper|scale|white bottle,4
+699,3495,3514,3504,gooseneck kettle|coffee dripper|scale|white bottle,4
+700,3500,3519,3509,gooseneck kettle|coffee dripper|scale|white bottle,4
+701,3505,3524,3514,gooseneck kettle|coffee dripper|scale|white bottle,4
+702,3510,3529,3519,gooseneck kettle|coffee dripper|scale|white bottle,4
+703,3515,3534,3524,gooseneck kettle|coffee dripper|scale|white bottle,4
+704,3520,3539,3529,gooseneck kettle|coffee dripper|scale|white bottle,4
+705,3525,3544,3534,gooseneck kettle|coffee dripper|scale|white bottle,4
+706,3530,3549,3539,gooseneck kettle|coffee dripper|scale|white bottle,4
+707,3535,3554,3544,gooseneck kettle|coffee dripper|scale|white bottle,4
+708,3540,3559,3549,gooseneck kettle|coffee dripper|scale|white bottle,4
+709,3545,3564,3554,gooseneck kettle|coffee dripper|scale|white bottle,4
+710,3550,3569,3559,gooseneck kettle|coffee dripper|scale|white bottle,4
+711,3555,3574,3564,gooseneck kettle|coffee dripper|scale|white bottle,4
+712,3560,3579,3569,gooseneck kettle|coffee dripper|scale|white bottle,4
+713,3565,3584,3574,gooseneck kettle|coffee dripper|scale|white bottle,4
+714,3570,3589,3579,gooseneck kettle|coffee dripper|scale|white bottle,4
+715,3575,3594,3584,gooseneck kettle|coffee dripper|scale|white bottle,4
+716,3580,3599,3589,gooseneck kettle|coffee dripper|scale|white bottle,4
+717,3585,3604,3594,gooseneck kettle|coffee dripper|scale|white bottle,4
+718,3590,3609,3599,gooseneck kettle|coffee dripper|scale|white bottle,4
+719,3595,3614,3604,gooseneck kettle|coffee dripper|scale|white bottle,4
+720,3600,3619,3609,gooseneck kettle|coffee dripper|scale|white bottle,4
+721,3605,3624,3614,gooseneck kettle|coffee dripper|scale|white bottle,4
+722,3610,3629,3619,gooseneck kettle|coffee dripper|scale|white bottle,4
+723,3615,3634,3624,gooseneck kettle|coffee dripper|scale|white bottle,4
+724,3620,3639,3629,gooseneck kettle|coffee dripper|scale|white bottle,4
+725,3625,3644,3634,gooseneck kettle|coffee dripper|scale|white bottle,4
+726,3630,3649,3639,gooseneck kettle|coffee dripper|scale|white bottle,4
+727,3635,3654,3644,gooseneck kettle|coffee dripper|scale|white bottle,4
+728,3640,3659,3649,gooseneck kettle|coffee dripper|scale|white bottle,4
+729,3645,3664,3654,gooseneck kettle|coffee dripper|scale|white bottle,4
+730,3650,3669,3659,gooseneck kettle|coffee dripper|scale|white bottle,4
+731,3655,3674,3664,gooseneck kettle|coffee dripper|scale|white bottle,4
+732,3660,3679,3669,gooseneck kettle|coffee dripper|scale|white bottle,4
+733,3665,3684,3674,gooseneck kettle|coffee dripper|scale|white bottle,4
+734,3670,3689,3679,gooseneck kettle|coffee dripper|scale|white bottle,4
+735,3675,3694,3684,gooseneck kettle|coffee dripper|scale|white bottle,4
+736,3680,3699,3689,gooseneck kettle|coffee dripper|scale|white bottle,4
+737,3685,3704,3694,gooseneck kettle|coffee dripper|scale|white bottle,4
+738,3690,3709,3699,gooseneck kettle|coffee dripper|scale|white bottle,4
+739,3695,3714,3704,gooseneck kettle|coffee dripper|scale|white bottle,4
+740,3700,3719,3709,gooseneck kettle|coffee dripper|scale|white bottle,4
+741,3705,3724,3714,gooseneck kettle|coffee dripper|scale|white bottle,4
+742,3710,3729,3719,gooseneck kettle|coffee dripper|scale|white bottle,4
+743,3715,3734,3724,gooseneck kettle|coffee dripper|scale|white bottle,4
+744,3720,3739,3729,gooseneck kettle|coffee dripper|scale|white bottle,4
+745,3725,3744,3734,gooseneck kettle|coffee dripper|scale|white bottle,4
+746,3730,3749,3739,gooseneck kettle|coffee dripper|scale|white bottle,4
+747,3735,3754,3744,gooseneck kettle|coffee dripper|scale|white bottle,4
+748,3740,3759,3749,gooseneck kettle|coffee dripper|scale|white bottle,4
+749,3745,3764,3754,gooseneck kettle|coffee dripper|scale|white bottle,4
+750,3750,3769,3759,gooseneck kettle|coffee dripper|scale|white bottle,4
+751,3755,3774,3764,gooseneck kettle|coffee dripper|scale|white bottle,4
+752,3760,3779,3769,gooseneck kettle|coffee dripper|scale|white bottle,4
+753,3765,3784,3774,gooseneck kettle|coffee dripper|scale|white bottle,4
+754,3770,3789,3779,gooseneck kettle|coffee dripper|scale|white bottle,4
+755,3775,3794,3784,gooseneck kettle|coffee dripper|scale|white bottle,4
+756,3780,3799,3789,gooseneck kettle|coffee dripper|scale|white bottle,4
+757,3785,3804,3794,gooseneck kettle|coffee dripper|scale|white bottle,4
+758,3790,3809,3799,gooseneck kettle|coffee dripper|scale|white bottle,4
+759,3795,3814,3804,gooseneck kettle|coffee dripper|scale|white bottle,4
+760,3800,3819,3809,white bottle|coffee dripper|scale,3
+761,3805,3824,3814,white bottle|coffee dripper|scale,3
+762,3810,3829,3819,white bottle|coffee dripper|scale,3
+763,3815,3834,3824,white bottle|coffee dripper|scale,3
+764,3820,3839,3829,white bottle|coffee dripper|scale,3
+765,3825,3844,3834,white bottle|coffee dripper|scale,3
+766,3830,3849,3839,white bottle|coffee dripper|scale,3
+767,3835,3854,3844,white bottle|coffee dripper|scale,3
+768,3840,3859,3849,white bottle|coffee dripper|scale,3
+769,3845,3864,3854,white bottle|coffee dripper|scale,3
+770,3850,3869,3859,white bottle|coffee dripper|scale,3
+771,3855,3874,3864,white bottle|coffee dripper|scale,3
+772,3860,3879,3869,white bottle|coffee dripper|scale,3
+773,3865,3884,3874,white bottle|coffee dripper|scale,3
+774,3870,3889,3879,white bottle|coffee dripper|scale,3
+775,3875,3894,3884,white bottle|coffee dripper|scale,3
+776,3880,3899,3889,white bottle|coffee dripper|scale,3
+777,3885,3904,3894,white bottle|coffee dripper|scale,3
+778,3890,3909,3899,white bottle|coffee dripper|scale,3
+779,3895,3914,3904,white bottle|coffee dripper|scale,3
+780,3900,3919,3909,white bottle|coffee dripper|scale,3
+781,3905,3924,3914,white bottle|coffee dripper|scale,3
+782,3910,3929,3919,white bottle|coffee dripper|scale,3
+783,3915,3934,3924,white bottle|coffee dripper|scale,3
+784,3920,3939,3929,white bottle,1
+785,3925,3944,3934,white bottle,1
+786,3930,3949,3939,white bottle,1
+787,3935,3954,3944,white bottle,1
+788,3940,3959,3949,white bottle,1
+789,3945,3964,3954,white bottle,1
+790,3950,3969,3959,white bottle,1
+791,3955,3974,3964,white bottle,1
+792,3960,3979,3969,white bottle,1
+793,3965,3984,3974,white bottle,1
+794,3970,3989,3979,white bottle,1
+795,3975,3994,3984,white bottle,1
+796,3980,3999,3989,white bottle,1
+797,3985,4004,3994,white bottle|coffee dripper|scale|mug,4
+798,3990,4009,3999,white bottle|coffee dripper|scale|mug,4
+799,3995,4014,4004,white bottle|coffee dripper|scale|mug,4
+800,4000,4019,4009,white bottle|coffee dripper|scale|mug,4
+801,4005,4024,4014,white bottle|coffee dripper|scale|mug,4
+802,4010,4029,4019,white bottle|coffee dripper|scale|mug,4
+803,4015,4034,4024,white bottle|coffee dripper|scale|mug,4
+804,4020,4039,4029,white bottle|coffee dripper|scale|mug,4
+805,4025,4044,4034,white bottle|coffee dripper|scale|mug,4
+806,4030,4049,4039,white bottle|coffee dripper|scale|mug,4
+807,4035,4054,4044,white bottle|coffee dripper|scale|mug,4
+808,4040,4059,4049,white bottle|coffee dripper|scale|mug,4
+809,4045,4064,4054,white bottle|coffee dripper|scale|mug,4
+810,4050,4069,4059,white bottle|coffee dripper|scale|mug,4
+811,4055,4074,4064,white bottle|coffee dripper|scale|mug,4
+812,4060,4079,4069,white bottle|coffee dripper|scale|mug,4
+813,4065,4084,4074,white bottle|coffee dripper|scale|mug,4
+814,4070,4089,4079,white bottle|coffee dripper|scale|mug,4
+815,4075,4094,4084,white bottle|coffee dripper|scale|mug,4
+816,4080,4099,4089,white bottle|coffee dripper|scale|mug,4
+817,4085,4104,4094,white bottle|coffee dripper|scale|mug,4
+818,4090,4109,4099,white bottle|coffee dripper|scale|mug,4
+819,4095,4114,4104,white bottle|coffee dripper|scale|mug,4
+820,4100,4119,4109,white bottle|coffee dripper|scale|mug,4
+821,4105,4124,4114,white bottle|coffee dripper|scale|mug,4
+822,4110,4129,4119,white bottle|coffee dripper|scale|mug,4
+823,4115,4134,4124,white bottle|coffee dripper|scale|mug,4
+824,4120,4139,4129,white bottle|coffee dripper|scale,3
+825,4125,4144,4134,white bottle|coffee dripper|scale,3
+826,4130,4149,4139,white bottle|coffee dripper|scale,3
+827,4135,4154,4144,white bottle|coffee dripper|scale,3
+828,4140,4159,4149,white bottle|coffee dripper|scale,3
+829,4145,4164,4154,white bottle|coffee dripper|scale,3
+830,4150,4169,4159,white bottle|coffee dripper|scale,3
+831,4155,4174,4164,white bottle|coffee dripper|scale,3
+832,4160,4179,4169,white bottle|coffee dripper|scale,3
+833,4165,4184,4174,white bottle|coffee dripper|scale,3
+834,4170,4189,4179,white bottle|coffee dripper|scale,3
+835,4175,4194,4184,white bottle|coffee dripper|scale,3
+836,4180,4199,4189,white bottle|coffee dripper|scale,3
+837,4185,4204,4194,white bottle|coffee dripper|scale,3
+838,4190,4209,4199,white bottle|coffee dripper|scale,3
+839,4195,4214,4204,white bottle|coffee dripper|scale,3
+840,4200,4219,4209,white bottle|coffee dripper|scale,3
+841,4205,4224,4214,white bottle|coffee dripper|scale,3
+842,4210,4229,4219,white bottle|coffee dripper|scale,3
+843,4215,4234,4224,white bottle|coffee dripper|scale,3
+844,4220,4239,4229,white bottle|coffee dripper|scale,3
+845,4225,4244,4234,white bottle|coffee dripper|scale,3
+846,4230,4249,4239,white bottle|coffee dripper|scale,3
+847,4235,4254,4244,white bottle|coffee dripper|scale,3
+848,4240,4259,4249,white bottle|coffee dripper|scale,3
+849,4245,4264,4254,white bottle|coffee dripper|scale,3
+850,4250,4269,4259,white bottle|coffee dripper|scale,3
+851,4255,4274,4264,white bottle|coffee dripper|scale,3
+852,4260,4279,4269,white bottle|coffee dripper|scale,3
+853,4265,4284,4274,white bottle|coffee dripper|scale,3
+854,4270,4289,4279,white bottle|coffee dripper|scale,3
+855,4275,4294,4284,white bottle|coffee dripper|scale,3
+856,4280,4299,4289,white bottle|coffee dripper|scale,3
+857,4285,4304,4294,white bottle|coffee dripper|scale,3
+858,4290,4309,4299,white bottle|coffee dripper|scale,3
+859,4295,4314,4304,white bottle|coffee dripper|scale,3
+860,4300,4319,4309,white bottle|coffee dripper|scale,3
+861,4305,4324,4314,white bottle|coffee dripper|scale,3
+862,4310,4329,4319,white bottle|coffee dripper|scale,3
+863,4315,4334,4324,white bottle|coffee dripper|scale,3
+864,4320,4339,4329,white bottle|coffee dripper|scale,3
+865,4325,4344,4334,white bottle|coffee dripper|scale,3
+866,4330,4349,4339,white bottle|coffee dripper|scale,3
+867,4335,4354,4344,white bottle|coffee dripper|scale,3
+868,4340,4359,4349,white bottle|coffee dripper|scale,3
+869,4345,4364,4354,white bottle|coffee dripper|scale,3
+870,4350,4369,4359,white bottle|coffee dripper|scale,3
+871,4355,4374,4364,white bottle|coffee dripper|scale,3
+872,4360,4379,4369,white bottle|coffee dripper|scale,3
+873,4365,4384,4374,white bottle|coffee dripper|scale,3
+874,4370,4389,4379,white bottle|coffee dripper|scale,3
+875,4375,4394,4384,white bottle|coffee dripper|scale,3
+876,4380,4399,4389,white bottle|coffee dripper|scale,3
+877,4385,4404,4394,white bottle|coffee dripper|scale|coffee equipment|small bottle|white mug|weighing scale,7
+878,4390,4409,4399,white bottle|coffee dripper|scale|coffee equipment|small bottle|white mug|weighing scale,7
+879,4395,4414,4404,white bottle|coffee dripper|scale|coffee equipment|small bottle|white mug|weighing scale,7
+880,4400,4419,4409,coffee equipment|small bottle|white mug|weighing scale,4
+881,4405,4424,4414,coffee equipment|small bottle|white mug|weighing scale,4
+882,4410,4429,4419,coffee equipment|small bottle|white mug|weighing scale,4
+883,4415,4434,4424,coffee equipment|small bottle|white mug|weighing scale,4
+884,4420,4439,4429,coffee equipment|small bottle|white mug|weighing scale,4
+885,4425,4444,4434,coffee equipment|small bottle|white mug|weighing scale,4
+886,4430,4449,4439,coffee equipment|small bottle|white mug|weighing scale,4
+887,4435,4454,4444,coffee equipment|small bottle|white mug|weighing scale,4
+888,4440,4459,4449,coffee equipment|small bottle|white mug|weighing scale,4
+889,4445,4464,4454,coffee equipment|small bottle|white mug|weighing scale,4
+890,4450,4469,4459,coffee equipment|small bottle|white mug|weighing scale,4
+891,4455,4474,4464,coffee equipment|small bottle|white mug|weighing scale,4
+892,4460,4479,4469,coffee equipment|small bottle|white mug|weighing scale,4
+893,4465,4484,4474,coffee equipment|small bottle|white mug|weighing scale,4
+894,4470,4489,4479,coffee equipment|small bottle|white mug|weighing scale,4
+895,4475,4494,4484,coffee equipment|small bottle|white mug|weighing scale,4
+896,4480,4499,4489,coffee equipment|small bottle|white mug|weighing scale,4
+897,4485,4504,4494,coffee equipment|small bottle|white mug|weighing scale,4
+898,4490,4509,4499,coffee equipment|small bottle|white mug|weighing scale,4
+899,4495,4514,4504,coffee equipment|small bottle|white mug|weighing scale,4
+900,4500,4519,4509,coffee equipment|small bottle|white mug|weighing scale,4
+901,4505,4524,4514,coffee equipment|small bottle|white mug|weighing scale,4
+902,4510,4529,4519,coffee equipment|small bottle|white mug|weighing scale,4
+903,4515,4534,4524,coffee equipment|small bottle|white mug|weighing scale,4
+904,4520,4539,4529,coffee equipment|small bottle|white mug|weighing scale,4
+905,4525,4544,4534,coffee equipment|small bottle|white mug|weighing scale,4
+906,4530,4549,4539,coffee equipment|small bottle|white mug|weighing scale,4
+907,4535,4554,4544,coffee equipment|small bottle|white mug|weighing scale,4
+908,4540,4559,4549,coffee equipment|small bottle|white mug|weighing scale,4
+909,4545,4564,4554,coffee equipment|small bottle|white mug|weighing scale,4
+910,4550,4569,4559,coffee equipment|small bottle|white mug|weighing scale,4
+911,4555,4574,4564,coffee equipment|small bottle|white mug|weighing scale,4
+912,4560,4579,4569,coffee equipment|small bottle|white mug|weighing scale,4
+913,4565,4584,4574,coffee equipment|small bottle|white mug|weighing scale,4
+914,4570,4589,4579,coffee equipment|small bottle|white mug|weighing scale,4
+915,4575,4594,4584,coffee equipment|small bottle|white mug|weighing scale,4
+916,4580,4599,4589,coffee equipment|small bottle|white mug|weighing scale,4
+917,4585,4604,4594,coffee equipment|small bottle|white mug|weighing scale,4
+918,4590,4609,4599,coffee equipment|small bottle|white mug|weighing scale,4
+919,4595,4614,4604,coffee equipment|small bottle|white mug|weighing scale,4
+920,4600,4619,4609,coffee equipment|small bottle|white mug|weighing scale,4
+921,4605,4624,4614,coffee equipment|small bottle|white mug|weighing scale,4
+922,4610,4629,4619,coffee equipment|small bottle|white mug|weighing scale,4
+923,4615,4634,4624,coffee equipment|small bottle|white mug|weighing scale,4
+924,4620,4639,4629,coffee equipment|small bottle|white mug|weighing scale,4
+925,4625,4644,4634,coffee equipment|small bottle|white mug|weighing scale,4
+926,4630,4649,4639,coffee equipment|small bottle|white mug|weighing scale,4
+927,4635,4654,4644,coffee equipment|small bottle|white mug|weighing scale,4
+928,4640,4659,4649,coffee equipment|small bottle|white mug|weighing scale,4
+929,4645,4664,4654,coffee equipment|small bottle|white mug|weighing scale,4
+930,4650,4669,4659,coffee equipment|small bottle|white mug|weighing scale,4
+931,4655,4674,4664,coffee equipment|small bottle|white mug|weighing scale,4
+932,4660,4679,4669,coffee equipment|small bottle|white mug|weighing scale,4
+933,4665,4684,4674,coffee equipment|small bottle|white mug|weighing scale,4
+934,4670,4689,4679,coffee equipment|small bottle|white mug|weighing scale,4
+935,4675,4694,4684,coffee equipment|small bottle|white mug|weighing scale,4
+936,4680,4699,4689,coffee equipment|small bottle|white mug|weighing scale,4
+937,4685,4704,4694,coffee equipment|small bottle|white mug|weighing scale,4
+938,4690,4709,4699,coffee equipment|small bottle|white mug|weighing scale,4
+939,4695,4714,4704,coffee equipment|small bottle|white mug|weighing scale,4
+940,4700,4719,4709,coffee equipment|small bottle|white mug|weighing scale,4
+941,4705,4724,4714,coffee equipment|small bottle|white mug|weighing scale,4
+942,4710,4729,4719,coffee equipment|small bottle|white mug|weighing scale,4
+943,4715,4734,4724,coffee equipment|small bottle|white mug|weighing scale,4
+944,4720,4739,4729,coffee equipment|small bottle|white mug|weighing scale,4
+945,4725,4744,4734,coffee equipment|small bottle|white mug|weighing scale,4
+946,4730,4749,4739,coffee equipment|small bottle|white mug|weighing scale,4
+947,4735,4754,4744,coffee equipment|small bottle|white mug|weighing scale,4
+948,4740,4759,4749,coffee equipment|small bottle|white mug|weighing scale,4
+949,4745,4764,4754,coffee equipment|small bottle|white mug|weighing scale,4
+950,4750,4769,4759,coffee equipment|small bottle|white mug|weighing scale,4
+951,4755,4774,4764,coffee equipment|small bottle|white mug|weighing scale,4
+952,4760,4779,4769,coffee equipment|small bottle|white mug|weighing scale,4
+953,4765,4784,4774,coffee equipment|small bottle|white mug|weighing scale,4
+954,4770,4789,4779,coffee equipment|small bottle|white mug|weighing scale,4
+955,4775,4794,4784,coffee equipment|small bottle|white mug|weighing scale,4
+956,4780,4799,4789,coffee equipment|small bottle|white mug|weighing scale,4
+957,4785,4804,4794,coffee equipment|small bottle|white mug|weighing scale|white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,9
+958,4790,4809,4799,coffee equipment|small bottle|white mug|weighing scale|white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,9
+959,4795,4814,4804,coffee equipment|small bottle|white mug|weighing scale|white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,9
+960,4800,4819,4809,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+961,4805,4824,4814,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+962,4810,4829,4819,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+963,4815,4834,4824,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+964,4820,4839,4829,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+965,4825,4844,4834,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+966,4830,4849,4839,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+967,4835,4854,4844,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+968,4840,4859,4849,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+969,4845,4864,4854,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+970,4850,4869,4859,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+971,4855,4874,4864,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+972,4860,4879,4869,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+973,4865,4884,4874,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+974,4870,4889,4879,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+975,4875,4894,4884,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+976,4880,4899,4889,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+977,4885,4904,4894,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+978,4890,4909,4899,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+979,4895,4914,4904,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+980,4900,4919,4909,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+981,4905,4924,4914,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+982,4910,4929,4919,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+983,4915,4934,4924,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+984,4920,4939,4929,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+985,4925,4944,4934,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+986,4930,4949,4939,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+987,4935,4954,4944,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+988,4940,4959,4949,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+989,4945,4964,4954,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+990,4950,4969,4959,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+991,4955,4974,4964,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+992,4960,4979,4969,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+993,4965,4984,4974,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+994,4970,4989,4979,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+995,4975,4994,4984,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+996,4980,4999,4989,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe,5
+997,4985,5004,4994,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe|carafe|coffee mug|scale,8
+998,4990,5009,4999,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe|carafe|coffee mug|scale,8
+999,4995,5014,5004,white coffee cup|digital scale with dripper|water bottle|metal pitcher|glass carafe|carafe|coffee mug|scale,8
+1000,5000,5019,5009,carafe|coffee mug|scale,3
+1001,5005,5024,5014,carafe|coffee mug|scale,3
+1002,5010,5029,5019,carafe|coffee mug|scale,3
+1003,5015,5034,5024,carafe|coffee mug|scale,3
+1004,5020,5039,5029,carafe|coffee mug|scale,3
+1005,5025,5044,5034,carafe|coffee mug|scale,3
+1006,5030,5049,5039,carafe|coffee mug|scale,3
+1007,5035,5054,5044,carafe|coffee mug|scale,3
+1008,5040,5059,5049,carafe|coffee mug|scale,3
+1009,5045,5064,5054,carafe|coffee mug|scale,3
+1010,5050,5069,5059,carafe|coffee mug|scale,3
+1011,5055,5074,5064,carafe|coffee mug|scale,3
+1012,5060,5079,5069,carafe|coffee mug|scale,3
+1013,5065,5084,5074,carafe|coffee mug|scale,3
+1014,5070,5089,5079,carafe|coffee mug|scale,3
+1015,5075,5094,5084,carafe|coffee mug|scale,3
+1016,5080,5099,5089,carafe|coffee mug|scale,3
+1017,5085,5104,5094,carafe|coffee mug|scale,3
+1018,5090,5109,5099,carafe|coffee mug|scale,3
+1019,5095,5114,5104,carafe|coffee mug|scale,3
+1020,5100,5119,5109,carafe|coffee mug|scale,3
+1021,5105,5124,5114,carafe|coffee mug|scale,3
+1022,5110,5129,5119,carafe|coffee mug|scale,3
+1023,5115,5134,5124,carafe|coffee mug|scale,3
+1024,5120,5139,5129,carafe|coffee mug|scale,3
+1025,5125,5144,5134,carafe|coffee mug|scale,3
+1026,5130,5149,5139,carafe|coffee mug|scale,3
+1027,5135,5154,5144,carafe|coffee mug|scale,3
+1028,5140,5159,5149,carafe|coffee mug|scale,3
+1029,5145,5164,5154,carafe|coffee mug|scale,3
+1030,5150,5169,5159,carafe|coffee mug|scale,3
+1031,5155,5174,5164,carafe|coffee mug|scale,3
+1032,5160,5179,5169,carafe|coffee mug|scale,3
+1033,5165,5184,5174,carafe|coffee mug|scale,3
+1034,5170,5189,5179,carafe|coffee mug|scale,3
+1035,5175,5194,5184,carafe|coffee mug|scale,3
+1036,5180,5199,5189,carafe|coffee mug|scale,3
+1037,5185,5204,5194,carafe|coffee mug|scale,3
+1038,5190,5209,5199,carafe|coffee mug|scale,3
+1039,5195,5214,5204,carafe|coffee mug|scale,3
+1040,5200,5219,5209,carafe|coffee mug|scale,3
+1041,5205,5224,5214,carafe|coffee mug|scale,3
+1042,5210,5229,5219,carafe|coffee mug|scale,3
+1043,5215,5234,5224,carafe|coffee mug|scale,3
+1044,5220,5239,5229,carafe|coffee mug|scale,3
+1045,5225,5244,5234,carafe|coffee mug|scale,3
+1046,5230,5249,5239,carafe|coffee mug|scale,3
+1047,5235,5254,5244,carafe|coffee mug|scale,3
+1048,5240,5259,5249,carafe|coffee mug|scale,3
+1049,5245,5264,5254,carafe|coffee mug|scale,3
+1050,5250,5269,5259,carafe|coffee mug|scale,3
+1051,5255,5274,5264,carafe|coffee mug|scale,3
+1052,5260,5279,5269,carafe|coffee mug|scale,3
+1053,5265,5284,5274,carafe|coffee mug|scale,3
+1054,5270,5289,5279,carafe|coffee mug|scale,3
+1055,5275,5294,5284,carafe|coffee mug|scale,3
+1056,5280,5299,5289,carafe|coffee mug|scale,3
+1057,5285,5304,5294,carafe|coffee mug|scale,3
+1058,5290,5309,5299,carafe|coffee mug|scale,3
+1059,5295,5314,5304,carafe|coffee mug|scale,3
+1060,5300,5319,5309,carafe|coffee mug|scale,3
+1061,5305,5324,5314,carafe|coffee mug|scale,3
+1062,5310,5329,5319,carafe|coffee mug|scale,3
+1063,5315,5334,5324,carafe|coffee mug|scale,3
+1064,5320,5339,5329,carafe|coffee mug|scale,3
+1065,5325,5344,5334,carafe|coffee mug|scale,3
+1066,5330,5349,5339,carafe|coffee mug|scale,3
+1067,5335,5354,5344,carafe|coffee mug|scale,3
+1068,5340,5359,5349,carafe|coffee mug|scale,3
+1069,5345,5364,5354,carafe|coffee mug|scale,3
+1070,5350,5369,5359,carafe|coffee mug|scale,3
+1071,5355,5374,5364,carafe|coffee mug|scale,3
+1072,5360,5379,5369,carafe|coffee mug|scale,3
+1073,5365,5384,5374,carafe|coffee mug|scale,3
+1074,5370,5389,5379,carafe|coffee mug|scale,3
+1075,5375,5394,5384,carafe|coffee mug|scale,3
+1076,5380,5399,5389,carafe|coffee mug|scale,3
+1077,5385,5404,5394,carafe|coffee mug|scale|milk pitcher|coffee cup|digital scale|bottle,7
+1078,5390,5409,5399,carafe|coffee mug|scale|milk pitcher|coffee cup|digital scale|bottle,7
+1079,5395,5414,5404,carafe|coffee mug|scale|milk pitcher|coffee cup|digital scale|bottle,7
+1080,5400,5419,5409,milk pitcher|coffee cup|digital scale|bottle,4
+1081,5405,5424,5414,milk pitcher|coffee cup|digital scale|bottle,4
+1082,5410,5429,5419,milk pitcher|coffee cup|digital scale|bottle,4
+1083,5415,5434,5424,milk pitcher|coffee cup|digital scale|bottle,4
+1084,5420,5439,5429,milk pitcher|coffee cup|digital scale|bottle,4
+1085,5425,5444,5434,milk pitcher|coffee cup|digital scale|bottle,4
+1086,5430,5449,5439,milk pitcher|coffee cup|digital scale|bottle,4
+1087,5435,5454,5444,milk pitcher|coffee cup|digital scale|bottle,4
+1088,5440,5459,5449,milk pitcher|coffee cup|digital scale|bottle,4
+1089,5445,5464,5454,milk pitcher|coffee cup|digital scale|bottle,4
+1090,5450,5469,5459,milk pitcher|coffee cup|digital scale|bottle,4
+1091,5455,5474,5464,milk pitcher|coffee cup|digital scale|bottle,4
+1092,5460,5479,5469,milk pitcher|coffee cup|digital scale|bottle,4
+1093,5465,5484,5474,milk pitcher|coffee cup|digital scale|bottle,4
+1094,5470,5489,5479,milk pitcher|coffee cup|digital scale|bottle,4
+1095,5475,5494,5484,milk pitcher|coffee cup|digital scale|bottle,4
+1096,5480,5499,5489,milk pitcher|coffee cup|digital scale|bottle,4
+1097,5485,5504,5494,milk pitcher|coffee cup|digital scale|bottle,4
+1098,5490,5509,5499,milk pitcher|coffee cup|digital scale|bottle,4
+1099,5495,5514,5504,milk pitcher|coffee cup|digital scale|bottle,4
+1100,5500,5519,5509,milk pitcher|coffee cup|digital scale|bottle,4
+1101,5505,5524,5514,milk pitcher|coffee cup|digital scale|bottle,4
+1102,5510,5529,5519,milk pitcher|coffee cup|digital scale|bottle,4
+1103,5515,5534,5524,milk pitcher|coffee cup|digital scale|bottle,4
+1104,5520,5539,5529,milk pitcher|coffee cup|digital scale|bottle,4
+1105,5525,5544,5534,milk pitcher|coffee cup|digital scale|bottle,4
+1106,5530,5549,5539,milk pitcher|coffee cup|digital scale|bottle,4
+1107,5535,5554,5544,milk pitcher|coffee cup|digital scale|bottle,4
+1108,5540,5559,5549,milk pitcher|coffee cup|digital scale|bottle,4
+1109,5545,5564,5554,milk pitcher|coffee cup|digital scale|bottle,4
+1110,5550,5569,5559,milk pitcher|coffee cup|digital scale|bottle,4
+1111,5555,5574,5564,milk pitcher|coffee cup|digital scale|bottle,4
+1112,5560,5579,5569,milk pitcher|coffee cup|digital scale|bottle,4
+1113,5565,5584,5574,milk pitcher|coffee cup|digital scale|bottle,4
+1114,5570,5589,5579,milk pitcher|coffee cup|digital scale|bottle,4
+1115,5575,5594,5584,milk pitcher|coffee cup|digital scale|bottle,4
+1116,5580,5599,5589,milk pitcher|coffee cup|digital scale|bottle,4
+1117,5585,5604,5594,milk pitcher|coffee cup|digital scale|bottle,4
+1118,5590,5609,5599,milk pitcher|coffee cup|digital scale|bottle,4
+1119,5595,5614,5604,milk pitcher|coffee cup|digital scale|bottle,4
+1120,5600,5619,5609,milk pitcher|coffee cup|digital scale|bottle,4
+1121,5605,5624,5614,milk pitcher|coffee cup|digital scale|bottle,4
+1122,5610,5629,5619,milk pitcher|coffee cup|digital scale|bottle,4
+1123,5615,5634,5624,milk pitcher|coffee cup|digital scale|bottle,4
+1124,5620,5639,5629,milk pitcher|coffee cup|digital scale|bottle,4
+1125,5625,5644,5634,milk pitcher|coffee cup|digital scale|bottle,4
+1126,5630,5649,5639,milk pitcher|coffee cup|digital scale|bottle,4
+1127,5635,5654,5644,milk pitcher|coffee cup|digital scale|bottle,4
+1128,5640,5659,5649,milk pitcher|coffee cup|digital scale|bottle,4
+1129,5645,5664,5654,milk pitcher|coffee cup|digital scale|bottle,4
+1130,5650,5669,5659,milk pitcher|coffee cup|digital scale|bottle,4
+1131,5655,5674,5664,milk pitcher|coffee cup|digital scale|bottle,4
+1132,5660,5679,5669,milk pitcher|coffee cup|digital scale|bottle,4
+1133,5665,5684,5674,milk pitcher|coffee cup|digital scale|bottle,4
+1134,5670,5689,5679,milk pitcher|coffee cup|digital scale|bottle,4
+1135,5675,5694,5684,milk pitcher|coffee cup|digital scale|bottle,4
+1136,5680,5699,5689,milk pitcher|coffee cup|digital scale|bottle,4
+1137,5685,5704,5694,milk pitcher|coffee cup|digital scale|bottle,4
+1138,5690,5709,5699,milk pitcher|coffee cup|digital scale|bottle,4
+1139,5695,5714,5704,milk pitcher|coffee cup|digital scale|bottle,4
+1140,5700,5719,5709,milk pitcher|coffee cup|digital scale|bottle,4
+1141,5705,5724,5714,milk pitcher|coffee cup|digital scale|bottle,4
+1142,5710,5729,5719,milk pitcher|coffee cup|digital scale|bottle,4
+1143,5715,5734,5724,milk pitcher|coffee cup|digital scale|bottle,4
+1144,5720,5739,5729,milk pitcher|coffee cup|digital scale|bottle,4
+1145,5725,5744,5734,milk pitcher|coffee cup|digital scale|bottle,4
+1146,5730,5749,5739,milk pitcher|coffee cup|digital scale|bottle,4
+1147,5735,5754,5744,milk pitcher|coffee cup|digital scale|bottle,4
+1148,5740,5759,5749,milk pitcher|coffee cup|digital scale|bottle,4
+1149,5745,5764,5754,milk pitcher|coffee cup|digital scale|bottle,4
+1150,5750,5769,5759,milk pitcher|coffee cup|digital scale|bottle,4
+1151,5755,5774,5764,milk pitcher|coffee cup|digital scale|bottle,4
+1152,5760,5779,5769,milk pitcher|coffee cup|digital scale|bottle,4
+1153,5765,5784,5774,milk pitcher|coffee cup|digital scale|bottle,4
+1154,5770,5789,5779,milk pitcher|coffee cup|digital scale|bottle,4
+1155,5775,5794,5784,milk pitcher|coffee cup|digital scale|bottle,4
+1156,5780,5799,5789,milk pitcher|coffee cup|digital scale|bottle,4
+1157,5785,5804,5794,milk pitcher|coffee cup|digital scale|bottle|coffee mug|stainless steel milk pitcher|table|milk bottle,8
+1158,5790,5809,5799,milk pitcher|coffee cup|digital scale|bottle|coffee mug|stainless steel milk pitcher|table|milk bottle,8
+1159,5795,5814,5804,milk pitcher|coffee cup|digital scale|bottle|coffee mug|stainless steel milk pitcher|table|milk bottle,8
+1160,5800,5819,5809,coffee mug|stainless steel milk pitcher|table|milk bottle|digital scale,5

results/single_episode_diagnostics/provenance.json

@@ -0,0 +1,142 @@
+{
+  "artifact_policy": "Only existing local artifacts are consumed. Missing labels/tasks are marked not_computed instead of filled.",
+  "source_suite_dir": "results/episode_task_suite",
+  "output_dir": "results/single_episode_diagnostics",
+  "shared_windows_shape": [
+    1161,
+    8378
+  ],
+  "starts_first_last": [
+    0,
+    5800
+  ],
+  "ends_first_last": [
+    19,
+    5819
+  ],
+  "feature_blocks": [
+    {
+      "name": "hand_left_joints",
+      "start": 0,
+      "end": 441,
+      "dim": 441
+    },
+    {
+      "name": "hand_right_joints",
+      "start": 441,
+      "end": 882,
+      "dim": 441
+    },
+    {
+      "name": "body_joints",
+      "start": 882,
+      "end": 1974,
+      "dim": 1092
+    },
+    {
+      "name": "body_contacts",
+      "start": 1974,
+      "end": 2121,
+      "dim": 147
+    },
+    {
+      "name": "camera_translation",
+      "start": 2121,
+      "end": 2142,
+      "dim": 21
+    },
+    {
+      "name": "camera_rotation_matrix",
+      "start": 2142,
+      "end": 2205,
+      "dim": 63
+    },
+    {
+      "name": "imu_accel_gyro",
+      "start": 2205,
+      "end": 2247,
+      "dim": 42
+    },
+    {
+      "name": "depth_confidence",
+      "start": 2247,
+      "end": 3227,
+      "dim": 980
+    },
+    {
+      "name": "video_fisheye_cam0",
+      "start": 3227,
+      "end": 3913,
+      "dim": 686
+    },
+    {
+      "name": "video_fisheye_cam1",
+      "start": 3913,
+      "end": 4599,
+      "dim": 686
+    },
+    {
+      "name": "video_fisheye_cam2",
+      "start": 4599,
+      "end": 5285,
+      "dim": 686
+    },
+    {
+      "name": "video_fisheye_cam3",
+      "start": 5285,
+      "end": 5971,
+      "dim": 686
+    },
+    {
+      "name": "video_stereo_left",
+      "start": 5971,
+      "end": 6657,
+      "dim": 686
+    },
+    {
+      "name": "video_stereo_right",
+      "start": 6657,
+      "end": 7343,
+      "dim": 686
+    },
+    {
+      "name": "caption_objects_interaction_text",
+      "start": 7343,
+      "end": 8239,
+      "dim": 896
+    },
+    {
+      "name": "slam_point_cloud",
+      "start": 8239,
+      "end": 8261,
+      "dim": 22
+    },
+    {
+      "name": "calibration",
+      "start": 8261,
+      "end": 8378,
+      "dim": 117
+    }
+  ],
+  "summary_report_core": {
+    "num_windows": 1161,
+    "feature_dim": 8378,
+    "window_frames": 20,
+    "stride_frames": 5,
+    "annotation": "data/sample/xperience-10m-sample/annotation.hdf5"
+  },
+  "input_file_hashes": {
+    "results/episode_task_suite/shared_windows.npz": "72ca8566186080580e7181ae08d5659dd19f2ad55a2a9613f5cee75467e501e6",
+    "results/episode_task_suite/windows.csv": "c075e43518a96934b68da97dceb45926c5fd71c846c8df5d94cc1a3d9e250739",
+    "results/episode_task_suite/feature_manifest.json": "a6662295f44c325490d3575edb1a0e88585554c57c3988263b78b59bec3970a0",
+    "results/episode_task_suite/summary_report.json": "04d2badfaa149cce10e218ea8e13745e620d062e4ec5fa9f209a3ed9926a2ef4",
+    "results/episode_task_suite/transition_detection/true_boundaries.csv": "5445144c216666d66e725be8313d7e2954670761a45d234df983acd94f7007f0",
+    "results/episode_task_suite/timeline_action/predictions.csv": "0ed609f2b5a2dc47b678a0b5d5fb6fc23d9a1cdd4f88dd85da8d8c9123c5565b",
+    "results/episode_task_suite/timeline_subtask/predictions.csv": "f12336affe6b907fc8b4b7655e4a16772e0cffb594789ebfc6a041bd7a251600",
+    "results/episode_task_suite/transition_detection/predictions.csv": "601a611bd9abf14bfac69d7fe44785b6567f4972fa6a80cb5d55f47b76c4d18c",
+    "results/episode_task_suite/next_action/predictions.csv": "1182cdeafe0282704d8b6a2c482a161d0bd97275a153e53d0501f3371ac61066",
+    "results/episode_task_suite/contact_prediction/predictions.csv": "161d7ddc9abc1d9cab08ce733f5f476c686a6a5046e81c9a8a4afbcf8064f43b",
+    "results/episode_task_suite/object_relevance/predictions.csv": "cd3a30437de815deafe67ac894e606c6fa87d0cccef15ae76757f9cf7cad5708",
+    "external_raw_sample/annotation.hdf5": "4a44b773c92715091c8d70e19f25e41a503160d2d6e7da98b47b99a538f9cad3"
+  }
+}

results/single_episode_diagnostics/timeline_overlay/TIMELINE_OVERLAY_REPORT.md

@@ -0,0 +1,17 @@
+# Timeline Prediction Overlay Report
+
+This report aligns existing prediction CSV files to the exported episode timeline. It does not rerun training.
+
+## Task-Level Correctness
+
+- Current Action Recognition: 10/343 correct (0.0292)
+- Current Subtask Recognition: 20/344 correct (0.0581)
+- Action Transition Detection: 322/348 correct (0.9253)
+- Next-Action Prediction: 12/348 correct (0.0345)
+- Contact State Prediction: 348/348 correct (1.0000)
+- Relevant Object Prediction: 2/348 correct (0.0057)
+
+## Files
+
+- `timeline_overlay.csv`: prediction rows with frame positions.
+- `timeline_overlay.svg`: visual overlay across the episode.

scripts/build_single_episode_explorer.py

@@ -0,0 +1,565 @@
+#!/usr/bin/env python3
+"""
+Build a static interactive explorer for the single Xperience-10M sample episode.
+
+The explorer is generated from committed/exported artifacts only. Raw MP4/HDF5
+files are not embedded or redistributed.
+"""
+
+from __future__ import annotations
+
+import argparse
+import csv
+import json
+from datetime import datetime, timezone
+from pathlib import Path
+
+import numpy as np
+
+
+TASK_DISPLAY = {
+    "timeline_action": "Current Action Recognition",
+    "timeline_subtask": "Current Subtask Recognition",
+    "transition_detection": "Action Transition Detection",
+    "next_action": "Next-Action Prediction",
+    "contact_prediction": "Contact State Prediction",
+    "object_relevance": "Relevant Object Prediction",
+}
+
+
+BLOCK_DISPLAY = {
+    "hand_left_joints": "Left Hand",
+    "hand_right_joints": "Right Hand",
+    "body_joints": "Body Joints",
+    "body_contacts": "Body Contacts",
+    "camera_translation": "Camera Translation",
+    "camera_rotation_matrix": "Camera Rotation",
+    "imu_accel_gyro": "IMU Accel/Gyro",
+    "depth_confidence": "Depth + Confidence",
+    "caption_objects_interaction_text": "Language Text",
+    "slam_point_cloud": "SLAM Point Cloud",
+    "calibration": "Calibration",
+}
+
+
+def parse_args() -> argparse.Namespace:
+    root = Path(__file__).resolve().parents[1]
+    parser = argparse.ArgumentParser(description="Build static single-episode explorer page.")
+    parser.add_argument("--workspace", type=Path, default=root)
+    parser.add_argument("--suite-dir", type=Path, default=root / "results/episode_task_suite")
+    parser.add_argument("--diagnostics-dir", type=Path, default=root / "results/single_episode_diagnostics")
+    parser.add_argument("--docs-dir", type=Path, default=root / "docs")
+    return parser.parse_args()
+
+
+def read_csv(path: Path) -> list[dict]:
+    with path.open(newline="", encoding="utf-8") as fp:
+        return list(csv.DictReader(fp))
+
+
+def read_json(path: Path):
+    return json.loads(path.read_text(encoding="utf-8"))
+
+
+def write_json(path: Path, data: dict) -> None:
+    path.parent.mkdir(parents=True, exist_ok=True)
+    path.write_text(json.dumps(data, indent=2, ensure_ascii=False), encoding="utf-8")
+
+
+def block_modality(name: str) -> str:
+    if name.startswith("video_"):
+        return "video"
+    if name.startswith("hand_") or name.startswith("body_"):
+        return "motion_capture"
+    if name.startswith("camera_") or name in {"slam_point_cloud", "calibration"}:
+        return "pose_slam"
+    if name.startswith("depth_"):
+        return "depth"
+    if name.startswith("imu_"):
+        return "inertial"
+    if name.startswith("caption_"):
+        return "language"
+    return "other"
+
+
+def load_predictions(suite_dir: Path) -> dict[str, dict[int, dict]]:
+    out: dict[str, dict[int, dict]] = {}
+    for task in TASK_DISPLAY:
+        path = suite_dir / task / "predictions.csv"
+        rows_by_window: dict[int, dict] = {}
+        if not path.exists():
+            out[task] = rows_by_window
+            continue
+        for row in read_csv(path):
+            if "window_index" not in row:
+                continue
+            idx = int(row["window_index"])
+            true_value = row.get("true_label") or row.get("true_objects") or row.get("true") or ""
+            pred_value = row.get("predicted_label") or row.get("predicted_objects") or row.get("predicted") or ""
+            if "correct" in row and row["correct"] != "":
+                correct = int(float(row["correct"]))
+            else:
+                correct = int(str(true_value) == str(pred_value))
+            rows_by_window[idx] = {
+                "true": true_value,
+                "predicted": pred_value,
+                "correct": correct,
+                "confidence": row.get("confidence", ""),
+            }
+        out[task] = rows_by_window
+    return out
+
+
+def build_action_segments(windows: list[dict]) -> list[dict]:
+    segments = []
+    if not windows:
+        return segments
+    current = windows[0]["action_label"]
+    start = int(windows[0]["start_frame"])
+    start_idx = int(windows[0]["window_index"])
+    last = windows[0]
+    for row in windows[1:]:
+        if row["action_label"] != current:
+            segments.append({
+                "action": current,
+                "start_frame": start,
+                "end_frame": int(last["end_frame"]),
+                "start_window": start_idx,
+                "end_window": int(last["window_index"]),
+            })
+            current = row["action_label"]
+            start = int(row["start_frame"])
+            start_idx = int(row["window_index"])
+        last = row
+    segments.append({
+        "action": current,
+        "start_frame": start,
+        "end_frame": int(last["end_frame"]),
+        "start_window": start_idx,
+        "end_window": int(last["window_index"]),
+    })
+    return segments
+
+
+def build_data(args: argparse.Namespace) -> dict:
+    suite_dir = args.suite_dir
+    diagnostics_dir = args.diagnostics_dir
+    windows = read_csv(suite_dir / "windows.csv")
+    manifest = read_json(suite_dir / "feature_manifest.json")
+    summary = read_json(suite_dir / "summary_report.json")
+    provenance = read_json(diagnostics_dir / "provenance.json")
+    object_rows = {int(r["window_index"]): r for r in read_csv(diagnostics_dir / "object_labels/window_object_labels.csv")}
+    ablation_rows = read_csv(diagnostics_dir / "modality_ablation/ablation_metrics.csv")
+    alignment_rows = read_csv(diagnostics_dir / "alignment_stress/alignment_shift_metrics.csv")
+    timeline_rows = read_csv(diagnostics_dir / "timeline_overlay/timeline_overlay.csv")
+    predictions = load_predictions(suite_dir)
+    X = np.load(suite_dir / "shared_windows.npz")["X"].astype(np.float32)
+
+    block_stats = {}
+    block_meta = []
+    for block in manifest:
+        name = block["name"]
+        start, end = int(block["start"]), int(block["end"])
+        values = X[:, start:end]
+        l2 = np.linalg.norm(values, axis=1)
+        mean_abs = np.mean(np.abs(values), axis=1)
+        max_l2 = float(max(l2.max(), 1e-8))
+        block_stats[name] = {
+            "l2": l2,
+            "mean_abs": mean_abs,
+            "relative": l2 / max_l2,
+        }
+        block_meta.append({
+            "name": name,
+            "display": BLOCK_DISPLAY.get(name, name.replace("_", " ").title()),
+            "modality": block_modality(name),
+            "start": start,
+            "end": end,
+            "dim": int(block["dim"]),
+        })
+
+    explorer_windows = []
+    for i, row in enumerate(windows):
+        idx = int(row["window_index"])
+        obj = object_rows.get(idx, {})
+        feature_stats = []
+        for block in block_meta:
+            s = block_stats[block["name"]]
+            feature_stats.append({
+                "name": block["name"],
+                "l2": round(float(s["l2"][i]), 6),
+                "mean_abs": round(float(s["mean_abs"][i]), 6),
+                "relative": round(float(s["relative"][i]), 6),
+            })
+        task_predictions = {}
+        for task, rows_by_window in predictions.items():
+            task_predictions[task] = rows_by_window.get(idx)
+        explorer_windows.append({
+            "window_index": idx,
+            "start_frame": int(row["start_frame"]),
+            "end_frame": int(row["end_frame"]),
+            "center_frame": int(row["center_frame"]),
+            "action": row["action_label"],
+            "subtask": row["subtask_label"],
+            "objects": [x for x in obj.get("objects", "").split("|") if x],
+            "feature_stats": feature_stats,
+            "predictions": task_predictions,
+        })
+
+    best_ablation = {}
+    for task in sorted({r["task"] for r in ablation_rows}):
+        computed = [r for r in ablation_rows if r["task"] == task and r["status"] == "computed" and r["score"]]
+        if not computed:
+            continue
+        best = max(computed, key=lambda r: float(r["score"]))
+        non_overlap = [r for r in computed if r.get("target_source_overlap") == "false"]
+        best_non_overlap = max(non_overlap, key=lambda r: float(r["score"])) if non_overlap else None
+        best_ablation[task] = {
+            "best": {
+                "modality_group": best["modality_group"],
+                "modality_display": best["modality_display"],
+                "score": float(best["score"]),
+                "primary_metric": best["primary_metric"],
+                "target_source_overlap": best["target_source_overlap"],
+            },
+            "best_non_overlap": None if best_non_overlap is None else {
+                "modality_group": best_non_overlap["modality_group"],
+                "modality_display": best_non_overlap["modality_display"],
+                "score": float(best_non_overlap["score"]),
+                "primary_metric": best_non_overlap["primary_metric"],
+            },
+        }
+
+    return {
+        "meta": {
+            "generated_at": datetime.now(timezone.utc).isoformat(),
+            "window_count": len(explorer_windows),
+            "feature_dim": int(X.shape[1]),
+            "object_label_rows": len(object_rows),
+            "object_vocab_count": len(read_json(diagnostics_dir / "object_labels/object_vocab.json")["vocab"]),
+            "timeline_prediction_rows": len(timeline_rows),
+            "source_policy": "Window-level labels, features, predictions, and diagnostics only. Raw Xperience-10M MP4/HDF5/RRD files are not embedded.",
+            "annotation_hash_recorded": any("annotation.hdf5" in key for key in provenance["input_file_hashes"]),
+            "summary": {
+                "num_windows": summary.get("num_windows"),
+                "feature_dim": summary.get("feature_dim"),
+                "window_frames": summary.get("window_frames"),
+                "stride_frames": summary.get("stride_frames"),
+            },
+        },
+        "tasks": TASK_DISPLAY,
+        "feature_blocks": block_meta,
+        "segments": build_action_segments(windows),
+        "windows": explorer_windows,
+        "ablation": {
+            "best_by_task": best_ablation,
+            "rows": ablation_rows,
+        },
+        "alignment": alignment_rows,
+    }
+
+
+HTML_TEMPLATE = """<!doctype html>
+<html lang="en">
+<head>
+  <meta charset="utf-8">
+  <meta name="viewport" content="width=device-width, initial-scale=1">
+  <title>Single-Episode Explorer | Ropedia Xperience-10M</title>
+  <meta name="description" content="Interactive window-level explorer for the Ropedia Xperience-10M single-episode diagnostics.">
+  <meta name="theme-color" content="#020502">
+  <link rel="icon" href="favicon.png" type="image/png" sizes="64x64">
+  <link rel="preconnect" href="https://fonts.googleapis.com">
+  <link rel="preconnect" href="https://fonts.gstatic.com" crossorigin>
+  <link href="https://fonts.googleapis.com/css2?family=Inter+Tight:wght@500;600;700;800&family=Space+Grotesk:wght@400;500;600;700&display=swap" rel="stylesheet">
+  <style>
+    :root { color-scheme: dark; --page:#020502; --panel:#071207; --surface:#0b1709; --ink:#f4f8ef; --muted:#a7b5a3; --line:rgba(164,242,127,.22); --soft:rgba(164,242,127,.12); --green:#a7f078; --cyan:#7ae5c3; --blue:#9bdfff; --red:#ff8f7a; --amber:#d8f4a5; --font-ui:"Inter Tight",system-ui,sans-serif; --font-copy:"Space Grotesk",system-ui,sans-serif; --max:1400px; }
+    * { box-sizing:border-box; }
+    body { margin:0; background: radial-gradient(circle at 80% 12%, rgba(164,242,127,.13), transparent 28%), radial-gradient(circle, rgba(164,242,127,.10) 1px, transparent 1.4px), var(--page); background-size:auto,22px 22px,auto; color:var(--ink); font-family:var(--font-copy); line-height:1.5; }
+    a { color:inherit; }
+    .wrap { width:min(var(--max), calc(100% - 42px)); margin:0 auto; }
+    header { position:sticky; top:0; z-index:10; background:rgba(2,5,2,.88); backdrop-filter:blur(16px); border-bottom:1px solid var(--soft); }
+    .nav { height:64px; display:flex; align-items:center; justify-content:space-between; gap:18px; }
+    .brand { display:flex; gap:11px; align-items:center; text-decoration:none; font-family:var(--font-ui); font-weight:760; }
+    .brand img { width:38px; height:38px; border:1px solid rgba(164,242,127,.42); border-radius:8px; background:#061006; }
+    .nav-links { display:flex; gap:14px; color:#c9d5c5; font-size:14px; }
+    .nav-links a { text-decoration:none; }
+    .hero { padding:54px 0 30px; border-bottom:1px solid var(--soft); }
+    h1 { max-width:900px; margin:0; font-family:var(--font-ui); font-size:clamp(42px, 6vw, 76px); line-height:.98; letter-spacing:0; }
+    .hero p { max-width:820px; margin:22px 0 0; color:#c7d1c3; font-size:18px; line-height:1.62; }
+    .stats { display:grid; grid-template-columns:repeat(4,minmax(0,1fr)); gap:10px; margin-top:28px; max-width:900px; }
+    .stat { border:1px solid var(--line); border-radius:8px; background:rgba(7,18,7,.82); padding:13px 14px; }
+    .stat strong { display:block; font-family:var(--font-ui); font-size:24px; line-height:1; }
+    .stat span { display:block; margin-top:6px; color:var(--muted); font-size:12px; }
+    main { padding:26px 0 70px; }
+    .shell { display:grid; grid-template-columns:330px minmax(0,1fr); gap:18px; align-items:start; }
+    .panel { border:1px solid var(--line); border-radius:8px; background:linear-gradient(180deg, rgba(164,242,127,.06), rgba(7,18,7,.88)); box-shadow:0 18px 48px rgba(0,0,0,.32); }
+    .side { position:sticky; top:84px; padding:18px; }
+    label { display:block; color:var(--muted); font-size:12px; font-family:var(--font-ui); font-weight:720; margin:14px 0 7px; }
+    input[type=range] { width:100%; accent-color:var(--green); }
+    select, input[type=search] { width:100%; min-height:40px; border:1px solid var(--soft); border-radius:6px; background:#020802; color:var(--ink); padding:9px 10px; font:inherit; }
+    .button-row { display:grid; grid-template-columns:1fr 1fr; gap:8px; margin-top:12px; }
+    button { border:1px solid var(--line); border-radius:6px; background:#10200d; color:var(--ink); min-height:38px; font:700 13px var(--font-ui); cursor:pointer; }
+    button:hover { border-color:var(--green); }
+    .timeline { padding:18px; margin-bottom:18px; }
+    .timeline-strip { position:relative; height:60px; border:1px solid var(--soft); border-radius:8px; overflow:hidden; background:#030803; }
+    .segment { position:absolute; top:0; bottom:0; border-right:1px solid rgba(2,5,2,.45); opacity:.92; }
+    .marker { position:absolute; top:0; bottom:0; width:3px; background:var(--ink); box-shadow:0 0 0 2px rgba(2,5,2,.9), 0 0 18px rgba(164,242,127,.6); }
+    .timeline-meta { display:flex; justify-content:space-between; gap:12px; margin-top:10px; color:var(--muted); font-size:12px; }
+    .content { display:grid; gap:18px; }
+    .window-panel { padding:22px; }
+    .window-head { display:grid; grid-template-columns:minmax(0,1fr) auto; gap:18px; align-items:start; border-bottom:1px solid var(--soft); padding-bottom:18px; }
+    h2 { margin:0; font-family:var(--font-ui); font-size:30px; line-height:1.1; }
+    .frame-pill { border:1px solid var(--line); border-radius:6px; padding:8px 10px; color:var(--green); font-family:var(--font-ui); font-size:13px; font-weight:760; white-space:nowrap; }
+    .subtle { color:var(--muted); }
+    .chips { display:flex; flex-wrap:wrap; gap:7px; margin-top:12px; }
+    .chip { border:1px solid var(--soft); background:rgba(164,242,127,.08); color:#e7f2df; border-radius:999px; padding:5px 8px; font-size:12px; }
+    .grid { display:grid; gap:12px; }
+    .pred-grid { grid-template-columns:repeat(3,minmax(0,1fr)); margin-top:18px; }
+    .pred { border:1px solid var(--soft); border-radius:8px; background:rgba(2,8,2,.72); padding:13px; min-height:118px; }
+    .pred h3 { margin:0 0 8px; font-family:var(--font-ui); font-size:15px; }
+    .pred p { margin:4px 0; color:#cdd8c8; font-size:13px; overflow-wrap:anywhere; }
+    .pred .ok { color:var(--green); font-weight:800; }
+    .pred .bad { color:var(--red); font-weight:800; }
+    .feature-grid { grid-template-columns:repeat(2,minmax(0,1fr)); }
+    .feature { display:grid; grid-template-columns:130px 1fr 62px; gap:10px; align-items:center; border-bottom:1px solid var(--soft); padding:10px 0; }
+    .feature:last-child { border-bottom:0; }
+    .feature-name { font-family:var(--font-ui); font-size:13px; color:#edf6e8; }
+    .bar { height:10px; border-radius:999px; background:rgba(164,242,127,.13); overflow:hidden; }
+    .bar span { display:block; height:100%; width:calc(var(--w) * 1%); background:linear-gradient(90deg,var(--cyan),var(--green)); }
+    .num { text-align:right; color:var(--muted); font-size:12px; font-variant-numeric:tabular-nums; }
+    .analysis-grid { display:grid; grid-template-columns:1fr 1fr; gap:18px; }
+    .analysis { padding:18px; }
+    .analysis h3 { margin:0 0 12px; font-family:var(--font-ui); font-size:20px; }
+    .rows { display:grid; gap:8px; }
+    .row { display:grid; grid-template-columns:1fr auto; gap:12px; border-bottom:1px solid var(--soft); padding:8px 0; color:#d8e4d3; font-size:13px; }
+    .row strong { color:var(--green); font-variant-numeric:tabular-nums; }
+    .note { margin-top:12px; color:var(--muted); font-size:12px; line-height:1.55; }
+    @media (max-width: 980px) { .shell,.analysis-grid { grid-template-columns:1fr; } .side { position:static; } .pred-grid,.feature-grid,.stats { grid-template-columns:1fr; } .window-head { grid-template-columns:1fr; } .nav-links { display:none; } }
+  </style>
+</head>
+<body>
+  <header>
+    <div class="wrap nav">
+      <a class="brand" href="index.html"><img src="assets/brand/xperience10m-logo-mark-192.png" alt=""><span>Ropedia Xperience-10M</span></a>
+      <nav class="nav-links"><a href="index.html">Project</a><a href="single_episode_explorer.html">Explorer</a><a href="data/single_episode_explorer.json">Data JSON</a></nav>
+    </div>
+  </header>
+  <section class="hero">
+    <div class="wrap">
+      <h1>Single-Episode Research Explorer</h1>
+      <p>Inspect the exported Xperience-10M sample windows, real object labels, model predictions, feature-block statistics, and diagnostic scores from one aligned episode.</p>
+      <div class="stats">
+        <div class="stat"><strong id="statWindows">-</strong><span>windows</span></div>
+        <div class="stat"><strong id="statDim">-</strong><span>feature dimensions</span></div>
+        <div class="stat"><strong id="statObjects">-</strong><span>object labels</span></div>
+        <div class="stat"><strong id="statPreds">-</strong><span>prediction rows</span></div>
+      </div>
+    </div>
+  </section>
+  <main>
+    <div class="wrap shell">
+      <aside class="panel side">
+        <label for="windowRange">Window</label>
+        <input id="windowRange" type="range" min="0" max="0" value="0">
+        <div class="button-row"><button id="prevWindow" type="button">Previous</button><button id="nextWindow" type="button">Next</button></div>
+        <label for="taskSelect">Task Focus</label>
+        <select id="taskSelect"></select>
+        <label for="searchBox">Search Action or Object</label>
+        <input id="searchBox" type="search" placeholder="e.g. Pour coffee, kettle">
+        <div class="button-row"><button id="firstMatch" type="button">First Match</button><button id="firstPred" type="button">First Predicted</button></div>
+        <p class="note">The page uses window-level exported artifacts only. Raw video, raw HDF5, and RRD assets are not embedded.</p>
+      </aside>
+      <section class="content">
+        <div class="panel timeline">
+          <div class="timeline-strip" id="timelineStrip"></div>
+          <div class="timeline-meta"><span id="timelineLeft"></span><span id="timelineRight"></span></div>
+        </div>
+        <section class="panel window-panel">
+          <div class="window-head">
+            <div>
+              <h2 id="windowTitle">Window</h2>
+              <p id="windowSubtitle" class="subtle"></p>
+              <div class="chips" id="objectChips"></div>
+            </div>
+            <div class="frame-pill" id="framePill"></div>
+          </div>
+          <div class="grid pred-grid" id="predictionGrid"></div>
+        </section>
+        <section class="analysis-grid">
+          <div class="panel analysis">
+            <h3>Feature Blocks</h3>
+            <div class="grid feature-grid" id="featureGrid"></div>
+          </div>
+          <div class="panel analysis">
+            <h3>Diagnostics</h3>
+            <div class="rows" id="diagnosticRows"></div>
+            <p class="note" id="diagnosticNote"></p>
+          </div>
+        </section>
+      </section>
+    </div>
+  </main>
+  <script id="explorer-data" type="application/json">__DATA__</script>
+  <script>
+    const DATA = JSON.parse(document.getElementById("explorer-data").textContent);
+    function hasPrediction(windowRecord, taskKey) {
+      return taskKey === "all" ? Object.values(windowRecord.predictions).some(Boolean) : Boolean(windowRecord.predictions[taskKey]);
+    }
+    function defaultWindowIndex() {
+      let best = 0;
+      let bestCount = -1;
+      DATA.windows.forEach((w) => {
+        const count = Object.values(w.predictions).filter(Boolean).length;
+        if (count > bestCount) { best = w.window_index; bestCount = count; }
+      });
+      return best;
+    }
+    const state = { index: defaultWindowIndex(), task: "all" };
+    const range = document.getElementById("windowRange");
+    const taskSelect = document.getElementById("taskSelect");
+    const searchBox = document.getElementById("searchBox");
+    const colors = ["#5ccf7d", "#7ae5c3", "#9bdfff", "#d8f4a5", "#f0a45e", "#cba8ff", "#ff8f7a"];
+    document.getElementById("statWindows").textContent = DATA.meta.window_count;
+    document.getElementById("statDim").textContent = DATA.meta.feature_dim;
+    document.getElementById("statObjects").textContent = DATA.meta.object_vocab_count;
+    document.getElementById("statPreds").textContent = DATA.meta.timeline_prediction_rows;
+    range.max = DATA.windows.length - 1;
+    for (const [key, label] of Object.entries(DATA.tasks)) {
+      const option = document.createElement("option");
+      option.value = key;
+      option.textContent = label;
+      taskSelect.appendChild(option);
+    }
+    const allOption = document.createElement("option");
+    allOption.value = "all";
+    allOption.textContent = "All Prediction Cards";
+    taskSelect.insertBefore(allOption, taskSelect.firstChild);
+    taskSelect.value = state.task;
+    function pct(value, min, max) { return ((value - min) / Math.max(1, max - min)) * 100; }
+    function splitObjects(value) { return String(value || "").split("|").filter(Boolean); }
+    function renderTimeline() {
+      const strip = document.getElementById("timelineStrip");
+      strip.innerHTML = "";
+      const minFrame = DATA.windows[0].start_frame;
+      const maxFrame = DATA.windows[DATA.windows.length - 1].end_frame;
+      DATA.segments.forEach((seg, i) => {
+        const el = document.createElement("div");
+        el.className = "segment";
+        el.style.left = pct(seg.start_frame, minFrame, maxFrame) + "%";
+        el.style.width = Math.max(0.3, pct(seg.end_frame, minFrame, maxFrame) - pct(seg.start_frame, minFrame, maxFrame)) + "%";
+        el.style.background = colors[i % colors.length];
+        el.title = `${seg.action} (${seg.start_frame}-${seg.end_frame})`;
+        el.addEventListener("click", () => { state.index = seg.start_window; render(); });
+        strip.appendChild(el);
+      });
+      const marker = document.createElement("div");
+      marker.className = "marker";
+      marker.style.left = pct(DATA.windows[state.index].center_frame, minFrame, maxFrame) + "%";
+      strip.appendChild(marker);
+      document.getElementById("timelineLeft").textContent = `frame ${minFrame}`;
+      document.getElementById("timelineRight").textContent = `frame ${maxFrame}`;
+    }
+    function renderPredictions(w) {
+      const grid = document.getElementById("predictionGrid");
+      grid.innerHTML = "";
+      const taskEntries = Object.entries(DATA.tasks).filter(([key]) => state.task === "all" || key === state.task);
+      for (const [key, label] of taskEntries) {
+        const pred = w.predictions[key];
+        const card = document.createElement("article");
+        card.className = "pred";
+        let body = "";
+        if (!pred) {
+          body = `<p class="subtle">No held-out prediction row for this window.</p>`;
+        } else {
+          const status = pred.correct ? `<span class="ok">correct</span>` : `<span class="bad">mismatch</span>`;
+          body = `<p>${status}</p><p><strong>true</strong>: ${escapeHtml(pred.true || "")}</p><p><strong>pred</strong>: ${escapeHtml(pred.predicted || "")}</p>`;
+          if (pred.confidence) body += `<p><strong>confidence</strong>: ${Number(pred.confidence).toFixed(3)}</p>`;
+        }
+        card.innerHTML = `<h3>${escapeHtml(label)}</h3>${body}`;
+        grid.appendChild(card);
+      }
+    }
+    function renderFeatures(w) {
+      const grid = document.getElementById("featureGrid");
+      grid.innerHTML = "";
+      for (const stat of w.feature_stats) {
+        const block = DATA.feature_blocks.find((b) => b.name === stat.name);
+        const row = document.createElement("div");
+        row.className = "feature";
+        row.innerHTML = `<span class="feature-name">${escapeHtml(block.display)}</span><span class="bar"><span style="--w:${Math.round(stat.relative * 100)}"></span></span><span class="num">${stat.l2.toFixed(2)}</span>`;
+        grid.appendChild(row);
+      }
+    }
+    function renderDiagnostics() {
+      const rows = document.getElementById("diagnosticRows");
+      rows.innerHTML = "";
+      const task = state.task === "all" ? "object_relevance" : state.task;
+      const diag = DATA.ablation.best_by_task[task];
+      if (diag) {
+        rows.innerHTML += `<div class="row"><span>Best modality for ${escapeHtml(DATA.tasks[task] || task)}</span><strong>${escapeHtml(diag.best.modality_display)} ${diag.best.score.toFixed(3)}</strong></div>`;
+        if (diag.best_non_overlap) rows.innerHTML += `<div class="row"><span>Best non-overlap modality</span><strong>${escapeHtml(diag.best_non_overlap.modality_display)} ${diag.best_non_overlap.score.toFixed(3)}</strong></div>`;
+      }
+      const zeroRows = DATA.alignment.filter((r) => Number(r.shift_windows) === 0);
+      zeroRows.slice(0, 5).forEach((r) => {
+        rows.innerHTML += `<div class="row"><span>${escapeHtml(r.query_display)} zero-shift retrieval MRR</span><strong>${Number(r.mrr).toFixed(3)}</strong></div>`;
+      });
+      document.getElementById("diagnosticNote").textContent = DATA.meta.source_policy;
+    }
+    function renderWindow() {
+      const w = DATA.windows[state.index];
+      range.value = state.index;
+      document.getElementById("windowTitle").textContent = `Window ${w.window_index}: ${w.action || "unlabeled action"}`;
+      document.getElementById("windowSubtitle").textContent = w.subtask || "No subtask label";
+      document.getElementById("framePill").textContent = `frames ${w.start_frame}-${w.end_frame}`;
+      const chips = document.getElementById("objectChips");
+      chips.innerHTML = "";
+      (w.objects.length ? w.objects : ["no object label"]).forEach((obj) => {
+        const chip = document.createElement("span");
+        chip.className = "chip";
+        chip.textContent = obj;
+        chips.appendChild(chip);
+      });
+      renderPredictions(w);
+      renderFeatures(w);
+      renderDiagnostics();
+    }
+    function render() { renderTimeline(); renderWindow(); }
+    function escapeHtml(s) { return String(s).replace(/[&<>"']/g, (c) => ({ "&":"&amp;", "<":"&lt;", ">":"&gt;", '"':"&quot;", "'":"&#39;" }[c])); }
+    range.addEventListener("input", () => { state.index = Number(range.value); render(); });
+    taskSelect.addEventListener("change", () => { state.task = taskSelect.value; render(); });
+    document.getElementById("prevWindow").addEventListener("click", () => { state.index = Math.max(0, state.index - 1); render(); });
+    document.getElementById("nextWindow").addEventListener("click", () => { state.index = Math.min(DATA.windows.length - 1, state.index + 1); render(); });
+    document.getElementById("firstPred").addEventListener("click", () => {
+      const found = DATA.windows.find((w) => hasPrediction(w, state.task));
+      if (found) { state.index = found.window_index; render(); }
+    });
+    document.getElementById("firstMatch").addEventListener("click", () => {
+      const q = searchBox.value.trim().toLowerCase();
+      if (!q) return;
+      const found = DATA.windows.find((w) => [w.action, w.subtask, ...w.objects].join(" ").toLowerCase().includes(q));
+      if (found) { state.index = found.window_index; render(); }
+    });
+    render();
+  </script>
+</body>
+</html>
+"""
+
+
+def write_html(path: Path, data: dict) -> None:
+    path.parent.mkdir(parents=True, exist_ok=True)
+    payload = json.dumps(data, ensure_ascii=False).replace("</script", "<\\/script")
+    path.write_text(HTML_TEMPLATE.replace("__DATA__", payload), encoding="utf-8")
+
+
+def main() -> None:
+    args = parse_args()
+    data = build_data(args)
+    write_json(args.docs_dir / "data/single_episode_explorer.json", data)
+    write_html(args.docs_dir / "single_episode_explorer.html", data)
+    print(f"Wrote {args.docs_dir / 'data/single_episode_explorer.json'}")
+    print(f"Wrote {args.docs_dir / 'single_episode_explorer.html'}")
+
+
+if __name__ == "__main__":
+    main()

scripts/single_episode_diagnostics.py

@@ -0,0 +1,1254 @@
+#!/usr/bin/env python3
+"""
+Single-episode diagnostics for the Xperience-10M task suite artifacts.
+
+This script is intentionally artifact-driven. It consumes the already exported
+one-episode shared feature table and prediction files, validates their shape and
+hashes, and writes diagnostics that can be manually traced back to those inputs.
+
+It does not invent labels or claim multi-episode generalization.
+"""
+
+from __future__ import annotations
+
+import argparse
+import csv
+import hashlib
+import html
+import json
+import math
+import sys
+from collections import OrderedDict
+from pathlib import Path
+from typing import Iterable
+
+import numpy as np
+
+
+TASKS = [
+    "timeline_action",
+    "timeline_subtask",
+    "transition_detection",
+    "next_action",
+    "hand_trajectory_forecast",
+    "contact_prediction",
+    "object_relevance",
+    "caption_grounding",
+    "cross_modal_retrieval",
+    "modality_reconstruction",
+    "temporal_order",
+    "misalignment_detection",
+]
+
+
+TASK_DISPLAY = {
+    "timeline_action": "Current Action Recognition",
+    "timeline_subtask": "Current Subtask Recognition",
+    "transition_detection": "Action Transition Detection",
+    "next_action": "Next-Action Prediction",
+    "hand_trajectory_forecast": "Future Hand Motion Forecasting",
+    "contact_prediction": "Contact State Prediction",
+    "object_relevance": "Relevant Object Prediction",
+    "caption_grounding": "Language-to-Time Grounding",
+    "cross_modal_retrieval": "Cross-Modal Window Retrieval",
+    "modality_reconstruction": "Sensor-to-Visual Reconstruction",
+    "temporal_order": "Temporal Order Verification",
+    "misalignment_detection": "Cross-Modal Misalignment Detection",
+}
+
+
+GROUP_DISPLAY = {
+    "all_features": "All Features",
+    "video": "Video",
+    "depth": "Depth",
+    "pose_slam": "Pose + SLAM",
+    "motion_capture": "Motion Capture",
+    "inertial": "Inertial",
+    "language": "Language",
+    "no_language": "All Except Language",
+    "motion_pose_inertial": "Motion + Pose + IMU",
+}
+
+
+def parse_args() -> argparse.Namespace:
+    workspace_default = Path(__file__).resolve().parents[1]
+    parser = argparse.ArgumentParser(description="Run single-episode diagnostics on real exported artifacts.")
+    parser.add_argument("--workspace", type=Path, default=workspace_default)
+    parser.add_argument(
+        "--suite-dir",
+        type=Path,
+        default=workspace_default / "results/episode_task_suite",
+        help="Existing single-episode task-suite artifact directory.",
+    )
+    parser.add_argument(
+        "--output-dir",
+        type=Path,
+        default=workspace_default / "results/single_episode_diagnostics",
+        help="Where to write new diagnostics. Existing task-suite outputs are not overwritten.",
+    )
+    parser.add_argument("--test-fraction", type=float, default=0.30)
+    parser.add_argument("--future-offset-windows", type=int, default=4)
+    parser.add_argument("--misalignment-shift-windows", type=int, default=8)
+    parser.add_argument("--ridge-l2", type=float, default=10.0)
+    parser.add_argument(
+        "--annotation",
+        type=Path,
+        default=None,
+        help="Optional raw annotation.hdf5. When provided, object relevance labels are exported from caption_frame_info_map.",
+    )
+    parser.add_argument(
+        "--homie-toolkit",
+        type=Path,
+        default=None,
+        help="Optional HOMIE-toolkit path. If omitted, inferred from --annotation when possible.",
+    )
+    return parser.parse_args()
+
+
+def write_json(path: Path, data: dict | list) -> None:
+    path.parent.mkdir(parents=True, exist_ok=True)
+    path.write_text(json.dumps(data, indent=2, ensure_ascii=False), encoding="utf-8")
+
+
+def write_text(path: Path, text: str) -> None:
+    path.parent.mkdir(parents=True, exist_ok=True)
+    path.write_text(text, encoding="utf-8")
+
+
+def write_csv(path: Path, rows: list[dict], fieldnames: list[str] | None = None) -> None:
+    path.parent.mkdir(parents=True, exist_ok=True)
+    if fieldnames is None:
+        keys: OrderedDict[str, None] = OrderedDict()
+        for row in rows:
+            for key in row:
+                keys.setdefault(key, None)
+        fieldnames = list(keys.keys())
+    with path.open("w", newline="", encoding="utf-8") as fp:
+        writer = csv.DictWriter(fp, fieldnames=fieldnames, lineterminator="\n")
+        writer.writeheader()
+        for row in rows:
+            writer.writerow({k: row.get(k, "") for k in fieldnames})
+
+
+def read_csv(path: Path) -> list[dict]:
+    with path.open(newline="", encoding="utf-8") as fp:
+        return list(csv.DictReader(fp))
+
+
+def sha256(path: Path) -> str:
+    h = hashlib.sha256()
+    with path.open("rb") as fp:
+        for chunk in iter(lambda: fp.read(1024 * 1024), b""):
+            h.update(chunk)
+    return h.hexdigest()
+
+
+def public_artifact_path(path: Path, repo_root: Path) -> str:
+    path = path.resolve()
+    try:
+        return str(path.relative_to(repo_root))
+    except ValueError:
+        pass
+    if path.name == "annotation.hdf5":
+        return "external_raw_sample/annotation.hdf5"
+    return path.name
+
+
+def public_source_reference(value: object) -> object:
+    if value in (None, ""):
+        return value
+    text = str(value)
+    if text.startswith("/") or "/" + "Users/" in text or "/" + "private/" in text:
+        path = Path(text)
+        if path.name == "annotation.hdf5":
+            return "external_raw_sample/annotation.hdf5"
+        return path.name
+    return text
+
+
+def load_inputs(suite_dir: Path) -> tuple[np.ndarray, np.ndarray, np.ndarray, list[dict], list[dict], dict]:
+    npz_path = suite_dir / "shared_windows.npz"
+    windows_path = suite_dir / "windows.csv"
+    manifest_path = suite_dir / "feature_manifest.json"
+    summary_path = suite_dir / "summary_report.json"
+    required = [npz_path, windows_path, manifest_path, summary_path]
+    missing = [str(p) for p in required if not p.exists()]
+    if missing:
+        raise FileNotFoundError(f"Missing required input artifacts: {missing}")
+
+    npz = np.load(npz_path)
+    X = np.asarray(npz["X"], dtype=np.float32)
+    starts = np.asarray(npz["starts"], dtype=np.int64)
+    ends = np.asarray(npz["ends"], dtype=np.int64)
+    windows = read_csv(windows_path)
+    manifest = json.loads(manifest_path.read_text(encoding="utf-8"))
+    summary = json.loads(summary_path.read_text(encoding="utf-8"))
+
+    if X.ndim != 2:
+        raise ValueError(f"Expected X to be 2-D, got shape {X.shape}")
+    if len(windows) != X.shape[0]:
+        raise ValueError(f"windows.csv rows ({len(windows)}) do not match X rows ({X.shape[0]})")
+    if len(starts) != X.shape[0] or len(ends) != X.shape[0]:
+        raise ValueError("starts/ends arrays do not match X rows")
+
+    for i, row in enumerate(windows):
+        if int(row["start_frame"]) != int(starts[i]) or int(row["end_frame"]) != int(ends[i]):
+            raise ValueError(f"Window start/end mismatch at row {i}")
+
+    cursor = 0
+    for block in manifest:
+        start, end, dim = int(block["start"]), int(block["end"]), int(block["dim"])
+        if start != cursor or end <= start or end - start != dim:
+            raise ValueError(f"Feature manifest has a gap, overlap, or bad dim at block {block}")
+        cursor = end
+    if cursor != X.shape[1]:
+        raise ValueError(f"Feature manifest ends at {cursor}, but X has {X.shape[1]} columns")
+
+    return X, starts, ends, windows, manifest, summary
+
+
+def chronological_split(n: int, test_fraction: float) -> tuple[np.ndarray, np.ndarray]:
+    if n < 2:
+        raise ValueError("Need at least two samples for a chronological split.")
+    split = int(round(n * (1.0 - test_fraction)))
+    split = max(1, min(split, n - 1))
+    return np.arange(split, dtype=np.int64), np.arange(split, n, dtype=np.int64)
+
+
+def block_indices(manifest: list[dict], include: Iterable[str] | None = None, exclude: Iterable[str] | None = None) -> np.ndarray:
+    include = list(include or [])
+    exclude = list(exclude or [])
+    idxs: list[int] = []
+    for block in manifest:
+        name = str(block["name"])
+        if include and not any(name == p or name.startswith(p) for p in include):
+            continue
+        if exclude and any(name == p or name.startswith(p) for p in exclude):
+            continue
+        idxs.extend(range(int(block["start"]), int(block["end"])))
+    return np.asarray(idxs, dtype=np.int64)
+
+
+def modality_groups(manifest: list[dict]) -> dict[str, np.ndarray]:
+    all_idx = block_indices(manifest)
+    language = block_indices(manifest, ["caption_objects_interaction_text"])
+    groups = {
+        "all_features": all_idx,
+        "video": block_indices(manifest, ["video_"]),
+        "depth": block_indices(manifest, ["depth_confidence"]),
+        "pose_slam": block_indices(manifest, ["camera_translation", "camera_rotation_matrix", "slam_point_cloud", "calibration"]),
+        "motion_capture": block_indices(manifest, ["hand_left_joints", "hand_right_joints", "body_joints", "body_contacts"]),
+        "inertial": block_indices(manifest, ["imu_accel_gyro"]),
+        "language": language,
+        "no_language": np.setdiff1d(all_idx, language),
+    }
+    return {name: idx for name, idx in groups.items() if len(idx) > 0}
+
+
+def encode_labels(labels: Iterable[str]) -> tuple[np.ndarray, list[str]]:
+    seen: OrderedDict[str, int] = OrderedDict()
+    encoded = []
+    for label in labels:
+        label = str(label)
+        if label not in seen:
+            seen[label] = len(seen)
+        encoded.append(seen[label])
+    return np.asarray(encoded, dtype=np.int64), list(seen.keys())
+
+
+def extract_objects(info: dict) -> list[str]:
+    objects = info.get("objects")
+    if isinstance(objects, list):
+        return [str(x).strip() for x in objects if str(x).strip()]
+    if objects:
+        return [str(objects).strip()]
+    return []
+
+
+def infer_homie_toolkit(annotation: Path, explicit: Path | None) -> Path | None:
+    if explicit is not None:
+        return explicit
+    annotation = annotation.resolve()
+    candidates = []
+    for parent in annotation.parents:
+        candidates.append(parent / "HOMIE-toolkit")
+    for candidate in candidates:
+        if candidate.exists():
+            return candidate
+    return None
+
+
+def load_object_targets_from_annotation(annotation: Path, windows: list[dict], toolkit: Path | None) -> dict:
+    annotation = annotation.resolve()
+    if not annotation.exists():
+        raise FileNotFoundError(annotation)
+    toolkit = infer_homie_toolkit(annotation, toolkit)
+    if toolkit is None or not toolkit.exists():
+        raise FileNotFoundError(f"HOMIE-toolkit not found for annotation {annotation}")
+    sys.path.insert(0, str(toolkit))
+    from data_loader import load_from_annotation_hdf5
+
+    ann = load_from_annotation_hdf5(annotation, 0, None, load_slam_point_cloud=False)
+    frame_info = ann["caption_frame_info_map"]
+    vocab: OrderedDict[str, int] = OrderedDict()
+    labels: list[list[str]] = []
+    rows_out: list[dict] = []
+    for row in windows:
+        counts: OrderedDict[str, int] = OrderedDict()
+        for frame in range(int(row["start_frame"]), int(row["end_frame"]) + 1):
+            for obj in extract_objects(frame_info.get(frame, {})):
+                counts[obj] = counts.get(obj, 0) + 1
+        objects = list(counts.keys())
+        for obj in objects:
+            if obj not in vocab:
+                vocab[obj] = len(vocab)
+        labels.append(objects)
+        rows_out.append({
+            "window_index": int(row["window_index"]),
+            "start_frame": int(row["start_frame"]),
+            "end_frame": int(row["end_frame"]),
+            "center_frame": int(row["center_frame"]),
+            "objects": "|".join(objects),
+            "object_count": int(len(objects)),
+        })
+    if not vocab:
+        raise ValueError("No object labels found in annotation caption_frame_info_map.")
+    Y = np.zeros((len(windows), len(vocab)), dtype=np.float32)
+    for i, objects in enumerate(labels):
+        for obj in objects:
+            Y[i, vocab[obj]] = 1.0
+    return {
+        "Y": Y,
+        "labels": labels,
+        "vocab": list(vocab.keys()),
+        "rows": rows_out,
+        "annotation": "external_raw_sample/annotation.hdf5",
+        "toolkit": "HOMIE-toolkit",
+        "source_note": (
+            "Object labels were exported from a raw Xperience-10M sample annotation. "
+            "The public artifact stores source type and hash instead of machine-specific file paths."
+        ),
+    }
+
+
+def standardize(train: np.ndarray, test: np.ndarray) -> tuple[np.ndarray, np.ndarray]:
+    mean = train.mean(axis=0, keepdims=True)
+    std = train.std(axis=0, keepdims=True)
+    std[std < 1e-6] = 1.0
+    return (train - mean) / std, (test - mean) / std
+
+
+def standardize_train_apply(train: np.ndarray, *arrays: np.ndarray) -> list[np.ndarray]:
+    mean = train.mean(axis=0, keepdims=True)
+    std = train.std(axis=0, keepdims=True)
+    std[std < 1e-6] = 1.0
+    return [(arr - mean) / std for arr in arrays]
+
+
+def ridge_predict(X_train: np.ndarray, Y_train: np.ndarray, X_test: np.ndarray, l2: float) -> np.ndarray:
+    X_train = np.asarray(X_train, dtype=np.float32)
+    X_test = np.asarray(X_test, dtype=np.float32)
+    Y_train = np.asarray(Y_train, dtype=np.float32)
+    Xb = np.concatenate([X_train, np.ones((X_train.shape[0], 1), dtype=np.float32)], axis=1)
+    Xtb = np.concatenate([X_test, np.ones((X_test.shape[0], 1), dtype=np.float32)], axis=1)
+    if Xb.shape[0] <= Xb.shape[1]:
+        K = Xb @ Xb.T
+        K.flat[:: K.shape[0] + 1] += l2
+        alpha = np.linalg.solve(K, Y_train)
+        return Xtb @ Xb.T @ alpha
+    A = Xb.T @ Xb
+    A.flat[:: A.shape[0] + 1] += l2
+    W = np.linalg.solve(A, Xb.T @ Y_train)
+    return Xtb @ W
+
+
+def classification_metrics(y_true: np.ndarray, y_pred: np.ndarray) -> dict:
+    classes = np.unique(np.concatenate([y_true, y_pred]))
+    f1s = []
+    recalls = []
+    for cls in classes:
+        tp = int(((y_true == cls) & (y_pred == cls)).sum())
+        fp = int(((y_true != cls) & (y_pred == cls)).sum())
+        fn = int(((y_true == cls) & (y_pred != cls)).sum())
+        precision = tp / (tp + fp) if tp + fp else 0.0
+        recall = tp / (tp + fn) if tp + fn else 0.0
+        f1 = 2 * precision * recall / (precision + recall) if precision + recall else 0.0
+        f1s.append(f1)
+        recalls.append(recall)
+    return {
+        "accuracy": float((y_true == y_pred).mean()) if len(y_true) else 0.0,
+        "macro_f1": float(np.mean(f1s)) if f1s else 0.0,
+        "balanced_accuracy": float(np.mean(recalls)) if recalls else 0.0,
+    }
+
+
+def multilabel_metrics(Y_true: np.ndarray, Y_pred: np.ndarray) -> dict:
+    Y_true = Y_true.astype(np.int64)
+    Y_pred = Y_pred.astype(np.int64)
+    tp = int(((Y_true == 1) & (Y_pred == 1)).sum())
+    fp = int(((Y_true == 0) & (Y_pred == 1)).sum())
+    fn = int(((Y_true == 1) & (Y_pred == 0)).sum())
+    precision = tp / (tp + fp) if tp + fp else 0.0
+    recall = tp / (tp + fn) if tp + fn else 0.0
+    micro_f1 = 2 * precision * recall / (precision + recall) if precision + recall else 0.0
+    per_f1 = []
+    for j in range(Y_true.shape[1]):
+        tpj = int(((Y_true[:, j] == 1) & (Y_pred[:, j] == 1)).sum())
+        fpj = int(((Y_true[:, j] == 0) & (Y_pred[:, j] == 1)).sum())
+        fnj = int(((Y_true[:, j] == 1) & (Y_pred[:, j] == 0)).sum())
+        pj = tpj / (tpj + fpj) if tpj + fpj else 0.0
+        rj = tpj / (tpj + fnj) if tpj + fnj else 0.0
+        per_f1.append(2 * pj * rj / (pj + rj) if pj + rj else 0.0)
+    return {
+        "micro_f1": float(micro_f1),
+        "macro_f1": float(np.mean(per_f1)) if per_f1 else 0.0,
+        "exact_match": float(np.mean(np.all(Y_true == Y_pred, axis=1))) if len(Y_true) else 0.0,
+        "precision": float(precision),
+        "recall": float(recall),
+    }
+
+
+def regression_metrics(y_true: np.ndarray, y_pred: np.ndarray) -> dict:
+    err = y_pred - y_true
+    mse = float(np.mean(err ** 2))
+    mae = float(np.mean(np.abs(err)))
+    denom = float(np.sum((y_true - y_true.mean(axis=0, keepdims=True)) ** 2))
+    r2 = 1.0 - float(np.sum(err ** 2)) / denom if denom > 1e-12 else 0.0
+    return {"mse": mse, "mae": mae, "r2": r2}
+
+
+def retrieval_metrics(pred_query: np.ndarray, target: np.ndarray) -> dict:
+    pred_query = pred_query.astype(np.float32)
+    target = target.astype(np.float32)
+    q_norm = pred_query / np.maximum(np.linalg.norm(pred_query, axis=1, keepdims=True), 1e-8)
+    t_norm = target / np.maximum(np.linalg.norm(target, axis=1, keepdims=True), 1e-8)
+    sims = q_norm @ t_norm.T
+    ranks = []
+    for i in range(sims.shape[0]):
+        order = np.argsort(-sims[i])
+        rank = int(np.where(order == i)[0][0]) + 1
+        ranks.append(rank)
+    ranks_arr = np.asarray(ranks, dtype=np.float32)
+    return {
+        "mrr": float(np.mean(1.0 / ranks_arr)) if len(ranks_arr) else 0.0,
+        "top1_accuracy": float(np.mean(ranks_arr <= 1)) if len(ranks_arr) else 0.0,
+        "top5_accuracy": float(np.mean(ranks_arr <= 5)) if len(ranks_arr) else 0.0,
+        "top10_accuracy": float(np.mean(ranks_arr <= 10)) if len(ranks_arr) else 0.0,
+        "median_rank": float(np.median(ranks_arr)) if len(ranks_arr) else 0.0,
+        "mean_rank": float(np.mean(ranks_arr)) if len(ranks_arr) else 0.0,
+        "num_queries": int(len(ranks_arr)),
+    }
+
+
+def onehot(y: np.ndarray, n_classes: int) -> np.ndarray:
+    out = np.zeros((len(y), n_classes), dtype=np.float32)
+    out[np.arange(len(y)), y] = 1.0
+    return out
+
+
+def fit_classification(
+    X: np.ndarray,
+    labels: np.ndarray,
+    train_idx: np.ndarray,
+    test_idx: np.ndarray,
+    l2: float,
+) -> tuple[dict, np.ndarray]:
+    y, class_names = encode_labels(labels)
+    train_classes = set(int(x) for x in y[train_idx])
+    test_classes = set(int(x) for x in y[test_idx])
+    unseen = [class_names[i] for i in sorted(test_classes - train_classes)]
+    X_train, X_test = standardize(X[train_idx], X[test_idx])
+    scores = ridge_predict(X_train, onehot(y[train_idx], len(class_names)), X_test, l2)
+    pred = scores.argmax(axis=1)
+    metrics = classification_metrics(y[test_idx], pred)
+    metrics.update({
+        "num_classes": len(class_names),
+        "num_train": int(len(train_idx)),
+        "num_test": int(len(test_idx)),
+        "unseen_test_classes": "|".join(unseen),
+        "unseen_test_class_count": int(len(unseen)),
+    })
+    return metrics, pred
+
+
+def fit_multilabel(
+    X: np.ndarray,
+    Y: np.ndarray,
+    train_idx: np.ndarray,
+    test_idx: np.ndarray,
+    l2: float,
+) -> tuple[dict, np.ndarray]:
+    X_train, X_test = standardize(X[train_idx], X[test_idx])
+    scores = ridge_predict(X_train, Y[train_idx], X_test, l2)
+    pred = (scores >= 0.5).astype(np.float32)
+    empty = np.where(pred.sum(axis=1) == 0)[0]
+    if len(empty):
+        pred[empty, np.argmax(scores[empty], axis=1)] = 1.0
+    metrics = multilabel_metrics(Y[test_idx], pred)
+    metrics.update({
+        "num_objects": int(Y.shape[1]),
+        "num_train": int(len(train_idx)),
+        "num_test": int(len(test_idx)),
+    })
+    return metrics, pred
+
+
+def frame_centers(windows: list[dict]) -> np.ndarray:
+    return np.asarray([int(row["center_frame"]) for row in windows], dtype=np.int64)
+
+
+def labels_from_windows(windows: list[dict], key: str) -> np.ndarray:
+    return np.asarray([str(row.get(key, "") or "") for row in windows], dtype=object)
+
+
+def transition_labels_from_boundaries(suite_dir: Path, centers: np.ndarray, tolerance_frames: int = 10) -> np.ndarray:
+    boundaries_path = suite_dir / "transition_detection/true_boundaries.csv"
+    if not boundaries_path.exists():
+        raise FileNotFoundError(boundaries_path)
+    rows = read_csv(boundaries_path)
+    boundary_frames = np.asarray([int(row.get("boundary_frame") or row.get("frame")) for row in rows], dtype=np.int64)
+    labels = np.zeros(len(centers), dtype=np.int64)
+    for i, center in enumerate(centers):
+        if len(boundary_frames) and np.min(np.abs(boundary_frames - center)) <= tolerance_frames:
+            labels[i] = 1
+    return np.asarray(["transition" if x else "steady" for x in labels], dtype=object)
+
+
+def task_target(
+    task: str,
+    X: np.ndarray,
+    windows: list[dict],
+    manifest: list[dict],
+    suite_dir: Path,
+    future_offset_windows: int,
+    object_targets: dict | None = None,
+) -> dict:
+    centers = frame_centers(windows)
+    n = len(windows)
+    train_idx, test_idx = chronological_split(n, 0.30)
+    all_idx = np.arange(n, dtype=np.int64)
+    if task == "timeline_action":
+        return {"kind": "classification", "labels": labels_from_windows(windows, "action_label"), "rows": all_idx}
+    if task == "timeline_subtask":
+        return {"kind": "classification", "labels": labels_from_windows(windows, "subtask_label"), "rows": all_idx}
+    if task == "transition_detection":
+        return {"kind": "classification", "labels": transition_labels_from_boundaries(suite_dir, centers), "rows": all_idx}
+    if task == "next_action":
+        rows = np.arange(0, n - future_offset_windows, dtype=np.int64)
+        labels = labels_from_windows(windows, "action_label")[rows + future_offset_windows]
+        return {"kind": "classification", "labels": labels, "rows": rows, "target_variant": "future action label from windows.csv"}
+    if task == "contact_prediction":
+        contacts = block_indices(manifest, ["body_contacts"])
+        rows = all_idx
+        labels = np.where(np.abs(X[:, contacts]).sum(axis=1) > 1e-8, "contact", "no_contact")
+        return {
+            "kind": "classification",
+            "labels": labels,
+            "rows": rows,
+            "target_source_blocks": "body_contacts",
+            "target_variant": "contact proxy derived from body_contacts feature block",
+        }
+    if task == "hand_trajectory_forecast":
+        rows = np.arange(0, n - future_offset_windows, dtype=np.int64)
+        hand = block_indices(manifest, ["hand_left_joints", "hand_right_joints"])
+        target = X[rows + future_offset_windows][:, hand]
+        return {
+            "kind": "regression",
+            "target": target,
+            "rows": rows,
+            "target_source_blocks": "future hand_left_joints|hand_right_joints",
+            "target_variant": "future hand feature vector from shared_windows.npz",
+        }
+    if task == "caption_grounding":
+        rows = all_idx
+        text = block_indices(manifest, ["caption_objects_interaction_text"])
+        return {"kind": "retrieval", "target": X[:, text], "rows": rows, "target_source_blocks": "caption_objects_interaction_text"}
+    if task in {"cross_modal_retrieval", "modality_reconstruction"}:
+        rows = all_idx
+        visual = block_indices(manifest, ["depth_confidence", "video_"])
+        return {"kind": "retrieval" if task == "cross_modal_retrieval" else "regression", "target": X[:, visual], "rows": rows, "target_source_blocks": "depth_confidence|video_*"}
+    if task == "temporal_order":
+        pairs = []
+        labels = []
+        for i in range(n - 1):
+            pairs.append((i, i + 1))
+            labels.append("forward")
+            pairs.append((i + 1, i))
+            labels.append("reversed")
+        return {"kind": "pair_classification", "pairs": np.asarray(pairs, dtype=np.int64), "labels": np.asarray(labels, dtype=object)}
+    if task == "misalignment_detection":
+        shift = 8
+        pairs = []
+        labels = []
+        for i in range(n - shift):
+            pairs.append((i, i))
+            labels.append("aligned")
+            pairs.append((i, i + shift))
+            labels.append("shifted")
+        return {"kind": "pair_classification", "pairs": np.asarray(pairs, dtype=np.int64), "labels": np.asarray(labels, dtype=object)}
+    if task == "object_relevance":
+        if object_targets is None:
+            return {
+                "kind": "not_available",
+                "reason": "raw annotation.hdf5 was not provided, so full-train object relevance labels could not be exported",
+            }
+        rows = all_idx
+        return {
+            "kind": "multilabel",
+            "target": object_targets["Y"],
+            "rows": rows,
+            "target_source_blocks": "caption_objects_interaction_text",
+            "target_variant": "object sets exported from annotation.hdf5 caption_frame_info_map",
+        }
+    raise KeyError(task)
+
+
+def target_overlap(group_idx: np.ndarray, target_info: dict, manifest: list[dict]) -> bool:
+    blocks = str(target_info.get("target_source_blocks", ""))
+    if not blocks:
+        return False
+    target_idx: list[int] = []
+    for part in blocks.split("|"):
+        part = part.strip()
+        if not part:
+            continue
+        prefix = part[:-1] if part.endswith("*") else part
+        target_idx.extend(block_indices(manifest, [prefix]).tolist())
+    if not target_idx:
+        return False
+    return bool(np.intersect1d(group_idx, np.asarray(target_idx, dtype=np.int64)).size)
+
+
+def pair_features(X: np.ndarray, pairs: np.ndarray, group_idx: np.ndarray, visual_idx: np.ndarray | None = None, task: str = "") -> np.ndarray:
+    left = X[pairs[:, 0]][:, group_idx]
+    right_source = X[pairs[:, 1]]
+    if task == "misalignment_detection" and visual_idx is not None:
+        right = right_source[:, visual_idx]
+    else:
+        right = right_source[:, group_idx]
+    diff = right[:, : min(left.shape[1], right.shape[1])] - left[:, : min(left.shape[1], right.shape[1])]
+    return np.concatenate([left, right, diff], axis=1).astype(np.float32)
+
+
+def run_modality_ablation(
+    X: np.ndarray,
+    windows: list[dict],
+    manifest: list[dict],
+    suite_dir: Path,
+    out_dir: Path,
+    args: argparse.Namespace,
+    object_targets: dict | None = None,
+) -> list[dict]:
+    groups = modality_groups(manifest)
+    visual_idx = block_indices(manifest, ["depth_confidence", "video_"])
+    contact_idx = block_indices(manifest, ["body_contacts"])
+    rows: list[dict] = []
+
+    for task in TASKS:
+        info = task_target(task, X, windows, manifest, suite_dir, args.future_offset_windows, object_targets)
+        for group_name, group_idx_raw in groups.items():
+            row = {
+                "task": task,
+                "task_display": TASK_DISPLAY[task],
+                "modality_group": group_name,
+                "modality_display": GROUP_DISPLAY[group_name],
+                "status": "computed",
+                "score": "",
+                "primary_metric": "",
+                "primary_metric_value": "",
+                "target_variant": info.get("target_variant", ""),
+                "target_source_overlap": "",
+                "reason": "",
+            }
+            if info["kind"] == "not_available":
+                row.update({"status": "not_computed", "reason": info["reason"]})
+                rows.append(row)
+                continue
+
+            group_idx = group_idx_raw
+            if task == "contact_prediction":
+                group_idx = np.setdiff1d(group_idx_raw, contact_idx)
+                if len(group_idx) == 0:
+                    row.update({"status": "not_computed", "reason": "input group would contain only contact target-source features"})
+                    rows.append(row)
+                    continue
+
+            try:
+                if info["kind"] == "classification":
+                    data_rows = np.asarray(info["rows"], dtype=np.int64)
+                    labels = np.asarray(info["labels"], dtype=object)
+                    train_local, test_local = chronological_split(len(data_rows), args.test_fraction)
+                    metrics, _ = fit_classification(
+                        X[data_rows][:, group_idx],
+                        labels,
+                        train_local,
+                        test_local,
+                        args.ridge_l2,
+                    )
+                    if metrics.get("status") == "not_computed":
+                        row.update(metrics)
+                    else:
+                        row.update(metrics)
+                        row["primary_metric"] = "macro_f1"
+                        row["primary_metric_value"] = metrics["macro_f1"]
+                        row["score"] = metrics["macro_f1"]
+                elif info["kind"] == "regression":
+                    data_rows = np.asarray(info["rows"], dtype=np.int64)
+                    target = np.asarray(info["target"], dtype=np.float32)
+                    train_local, test_local = chronological_split(len(data_rows), args.test_fraction)
+                    Xin_train, Xin_test = standardize(X[data_rows[train_local]][:, group_idx], X[data_rows[test_local]][:, group_idx])
+                    Y_train, Y_test = standardize(info["target"][train_local], info["target"][test_local])
+                    pred = ridge_predict(Xin_train, Y_train, Xin_test, args.ridge_l2)
+                    metrics = regression_metrics(Y_test, pred)
+                    row.update(metrics)
+                    row["primary_metric"] = "mae"
+                    row["primary_metric_value"] = metrics["mae"]
+                    row["score"] = 1.0 / (1.0 + metrics["mae"])
+                elif info["kind"] == "multilabel":
+                    data_rows = np.asarray(info["rows"], dtype=np.int64)
+                    target = np.asarray(info["target"], dtype=np.float32)
+                    train_local, test_local = chronological_split(len(data_rows), args.test_fraction)
+                    metrics, _ = fit_multilabel(X[data_rows][:, group_idx], target, train_local, test_local, args.ridge_l2)
+                    row.update(metrics)
+                    row["primary_metric"] = "micro_f1"
+                    row["primary_metric_value"] = metrics["micro_f1"]
+                    row["score"] = metrics["micro_f1"]
+                elif info["kind"] == "retrieval":
+                    data_rows = np.asarray(info["rows"], dtype=np.int64)
+                    target = np.asarray(info["target"], dtype=np.float32)
+                    train_local, test_local = chronological_split(len(data_rows), args.test_fraction)
+                    Xin_train, Xin_test = standardize(X[data_rows[train_local]][:, group_idx], X[data_rows[test_local]][:, group_idx])
+                    Y_train, Y_test = standardize(target[train_local], target[test_local])
+                    pred = ridge_predict(Xin_train, Y_train, Xin_test, args.ridge_l2)
+                    metrics = retrieval_metrics(pred, Y_test)
+                    row.update(metrics)
+                    row["primary_metric"] = "mrr"
+                    row["primary_metric_value"] = metrics["mrr"]
+                    row["score"] = metrics["mrr"]
+                elif info["kind"] == "pair_classification":
+                    pairs = np.asarray(info["pairs"], dtype=np.int64)
+                    labels = np.asarray(info["labels"], dtype=object)
+                    train_local, test_local = chronological_split(len(pairs), args.test_fraction)
+                    feats = pair_features(X, pairs, group_idx, visual_idx=visual_idx, task=task)
+                    metrics, _ = fit_classification(feats, labels, train_local, test_local, args.ridge_l2)
+                    if metrics.get("status") == "not_computed":
+                        row.update(metrics)
+                    else:
+                        row.update(metrics)
+                        row["primary_metric"] = "macro_f1"
+                        row["primary_metric_value"] = metrics["macro_f1"]
+                        row["score"] = metrics["macro_f1"]
+                else:
+                    raise ValueError(info["kind"])
+                row["target_source_overlap"] = str(target_overlap(group_idx, info, manifest)).lower()
+            except Exception as exc:  # keep the matrix auditable instead of silently dropping failures
+                row.update({"status": "not_computed", "reason": f"{type(exc).__name__}: {exc}"})
+            rows.append(row)
+
+    write_csv(out_dir / "modality_ablation/ablation_metrics.csv", rows)
+    write_json(
+        out_dir / "modality_ablation/ablation_summary.json",
+        {
+            "description": "Compact ridge-head ablation over real shared_windows.npz feature blocks.",
+            "num_rows": len(rows),
+            "num_computed": sum(1 for r in rows if r.get("status") == "computed"),
+            "num_not_computed": sum(1 for r in rows if r.get("status") != "computed"),
+            "groups": {name: int(len(idx)) for name, idx in groups.items()},
+            "tasks": TASKS,
+            "object_relevance_labels": "annotation.hdf5" if object_targets is not None else "not_available",
+        },
+    )
+    render_ablation_svg(rows, out_dir / "modality_ablation/ablation_matrix.svg")
+    write_modality_report(rows, out_dir / "modality_ablation/MODALITY_ABLATION_REPORT.md")
+    return rows
+
+
+def color_for_score(score: float | None) -> str:
+    if score is None or math.isnan(score):
+        return "#20251f"
+    score = max(0.0, min(1.0, score))
+    r = int(37 + (154 - 37) * (1.0 - score))
+    g = int(72 + (224 - 72) * score)
+    b = int(54 + (101 - 54) * score)
+    return f"#{r:02x}{g:02x}{b:02x}"
+
+
+def render_ablation_svg(rows: list[dict], path: Path) -> None:
+    groups = list(GROUP_DISPLAY.keys())
+    tasks = TASKS
+    cell_w, cell_h = 132, 34
+    left, top = 300, 98
+    width = left + cell_w * len(groups) + 44
+    height = top + cell_h * len(tasks) + 86
+    by_key = {(r["task"], r["modality_group"]): r for r in rows}
+    parts = [
+        f'<svg xmlns="http://www.w3.org/2000/svg" width="{width}" height="{height}" viewBox="0 0 {width} {height}">',
+        '<rect width="100%" height="100%" fill="#10160f"/>',
+        '<text x="28" y="38" fill="#eef5e8" font-family="Inter, Arial" font-size="24" font-weight="700">Single-Episode Modality Ablation Matrix</text>',
+        '<text x="28" y="66" fill="#a7b5a3" font-family="Inter, Arial" font-size="13">Scores are recomputed from shared_windows.npz; gray cells are intentionally not computed.</text>',
+    ]
+    for j, group in enumerate(groups):
+        x = left + j * cell_w + 6
+        parts.append(
+            f'<text x="{x}" y="{top - 18}" fill="#cbd8c8" font-family="Inter, Arial" font-size="12" transform="rotate(-25 {x} {top - 18})">{html.escape(GROUP_DISPLAY[group])}</text>'
+        )
+    for i, task in enumerate(tasks):
+        y = top + i * cell_h
+        parts.append(f'<text x="28" y="{y + 22}" fill="#e7efe2" font-family="Inter, Arial" font-size="13">{html.escape(TASK_DISPLAY[task])}</text>')
+        for j, group in enumerate(groups):
+            x = left + j * cell_w
+            row = by_key.get((task, group), {})
+            if row.get("status") == "computed" and row.get("score") != "":
+                score = float(row["score"])
+                fill = color_for_score(score)
+                label = f'{score:.2f}'
+                label_fill = "#061006" if score > 0.62 else "#edf5e7"
+            else:
+                fill = "#222720"
+                label = "n/a"
+                label_fill = "#7b8678"
+            parts.append(f'<rect x="{x}" y="{y}" width="{cell_w - 8}" height="{cell_h - 7}" rx="5" fill="{fill}" stroke="#34402f" stroke-width="1"/>')
+            parts.append(f'<text x="{x + (cell_w - 8) / 2}" y="{y + 19}" text-anchor="middle" fill="{label_fill}" font-family="Inter, Arial" font-size="12" font-weight="700">{label}</text>')
+    parts.extend(
+        [
+            f'<text x="28" y="{height - 34}" fill="#a7b5a3" font-family="Inter, Arial" font-size="12">Metric: macro-F1 / MRR / 1/(1+MAE), depending on task type. See CSV for raw values and overlap flags.</text>',
+            "</svg>",
+        ]
+    )
+    write_text(path, "\n".join(parts))
+
+
+def write_modality_report(rows: list[dict], path: Path) -> None:
+    computed = [r for r in rows if r.get("status") == "computed" and r.get("score") != ""]
+    by_task: dict[str, list[dict]] = {t: [] for t in TASKS}
+    for row in computed:
+        by_task[row["task"]].append(row)
+    lines = [
+        "# Single-Episode Modality Ablation Report",
+        "",
+        "This diagnostic reruns compact ridge heads on the exported one-episode feature matrix. It is useful for checking which real feature blocks can support each task on this episode, not for estimating dataset-wide generalization.",
+        "",
+        "No synthetic labels are introduced. Derived proxy targets are marked in `target_variant`, and feature groups that overlap with the target source are marked in `target_source_overlap`.",
+        "",
+        "## Best Computed Group Per Task",
+        "",
+    ]
+    for task in TASKS:
+        task_rows = by_task.get(task, [])
+        if not task_rows:
+            reasons = sorted({r.get("reason", "") for r in rows if r["task"] == task and r.get("reason")})
+            lines.append(f"- {TASK_DISPLAY[task]}: not computed ({'; '.join(reasons)})")
+            continue
+        best = max(task_rows, key=lambda r: float(r["score"]))
+        line = (
+            f"- {TASK_DISPLAY[task]}: {best['modality_display']} score={float(best['score']):.4f}, "
+            f"{best['primary_metric']}={float(best['primary_metric_value']):.4f}, target overlap={best['target_source_overlap']}"
+        )
+        if best["target_source_overlap"] == "true":
+            no_overlap = [r for r in task_rows if r.get("target_source_overlap") == "false"]
+            if no_overlap:
+                alt = max(no_overlap, key=lambda r: float(r["score"]))
+                line += (
+                    f"; best non-overlap: {alt['modality_display']} score={float(alt['score']):.4f}, "
+                    f"{alt['primary_metric']}={float(alt['primary_metric_value']):.4f}"
+                )
+        lines.append(line)
+    lines.extend(
+        [
+            "",
+            "## Files",
+            "",
+            "- `ablation_metrics.csv`: every task/modality pair, including not-computed rows and reasons.",
+            "- `ablation_matrix.svg`: compact heatmap for manual inspection.",
+            "- `ablation_summary.json`: group dimensions and computed/not-computed counts.",
+        ]
+    )
+    write_text(path, "\n".join(lines) + "\n")
+
+
+def run_timeline_overlay(suite_dir: Path, windows: list[dict], out_dir: Path) -> list[dict]:
+    overlay_tasks = [
+        "timeline_action",
+        "timeline_subtask",
+        "transition_detection",
+        "next_action",
+        "contact_prediction",
+        "object_relevance",
+    ]
+    window_by_index = {int(row["window_index"]): row for row in windows}
+    rows: list[dict] = []
+    for task in overlay_tasks:
+        pred_path = suite_dir / task / "predictions.csv"
+        if not pred_path.exists():
+            rows.append({"task": task, "status": "not_available", "reason": f"missing {pred_path}"})
+            continue
+        for pred in read_csv(pred_path):
+            try:
+                idx = int(pred["window_index"])
+            except KeyError:
+                continue
+            window = window_by_index.get(idx, {})
+            true_value = pred.get("true_label") or pred.get("true_objects") or pred.get("true") or ""
+            pred_value = pred.get("predicted_label") or pred.get("predicted_objects") or pred.get("predicted") or ""
+            if "correct" in pred and pred["correct"] != "":
+                correct = int(float(pred["correct"]))
+            else:
+                correct = int(str(true_value) == str(pred_value))
+            rows.append(
+                {
+                    "task": task,
+                    "task_display": TASK_DISPLAY[task],
+                    "status": "observed_prediction",
+                    "window_index": idx,
+                    "start_frame": pred.get("start_frame") or window.get("start_frame", ""),
+                    "end_frame": pred.get("end_frame") or window.get("end_frame", ""),
+                    "center_frame": pred.get("center_frame") or window.get("center_frame", ""),
+                    "true_value": true_value,
+                    "predicted_value": pred_value,
+                    "confidence": pred.get("confidence", ""),
+                    "correct": correct,
+                }
+            )
+
+    write_csv(out_dir / "timeline_overlay/timeline_overlay.csv", rows)
+    render_timeline_svg(rows, windows, suite_dir, out_dir / "timeline_overlay/timeline_overlay.svg")
+    write_timeline_report(rows, out_dir / "timeline_overlay/TIMELINE_OVERLAY_REPORT.md")
+    return rows
+
+
+def render_timeline_svg(rows: list[dict], windows: list[dict], suite_dir: Path, path: Path) -> None:
+    tasks = ["timeline_action", "timeline_subtask", "transition_detection", "next_action", "contact_prediction", "object_relevance"]
+    min_frame = min(int(r["start_frame"]) for r in windows)
+    max_frame = max(int(r["end_frame"]) for r in windows)
+    left, top = 260, 84
+    row_h, plot_w = 48, 1100
+    width = left + plot_w + 38
+    height = top + row_h * len(tasks) + 92
+    by_task: dict[str, list[dict]] = {t: [] for t in tasks}
+    for row in rows:
+        if row.get("status") == "observed_prediction":
+            by_task[row["task"]].append(row)
+
+    def x_for(frame: int) -> float:
+        return left + (frame - min_frame) / max(1, max_frame - min_frame) * plot_w
+
+    parts = [
+        f'<svg xmlns="http://www.w3.org/2000/svg" width="{width}" height="{height}" viewBox="0 0 {width} {height}">',
+        '<rect width="100%" height="100%" fill="#10160f"/>',
+        '<text x="28" y="38" fill="#eef5e8" font-family="Inter, Arial" font-size="24" font-weight="700">Held-Out Timeline Prediction Overlay</text>',
+        '<text x="28" y="64" fill="#a7b5a3" font-family="Inter, Arial" font-size="13">Bars are existing real prediction rows aligned back to the exported episode timeline.</text>',
+    ]
+    boundaries_path = suite_dir / "transition_detection/true_boundaries.csv"
+    boundary_frames = []
+    if boundaries_path.exists():
+        boundary_frames = [int(r.get("boundary_frame") or r.get("frame")) for r in read_csv(boundaries_path)]
+    for i, task in enumerate(tasks):
+        y = top + i * row_h
+        parts.append(f'<text x="28" y="{y + 25}" fill="#e7efe2" font-family="Inter, Arial" font-size="13">{html.escape(TASK_DISPLAY[task])}</text>')
+        parts.append(f'<line x1="{left}" y1="{y + 20}" x2="{left + plot_w}" y2="{y + 20}" stroke="#2a3428" stroke-width="18" stroke-linecap="round"/>')
+        for row in by_task[task]:
+            if not row.get("start_frame") or not row.get("end_frame"):
+                continue
+            x1 = x_for(int(float(row["start_frame"])))
+            x2 = max(x1 + 2, x_for(int(float(row["end_frame"]))))
+            fill = "#8ee06a" if int(row["correct"]) else "#e46b5f"
+            parts.append(f'<rect x="{x1:.2f}" y="{y + 11}" width="{x2 - x1:.2f}" height="18" rx="3" fill="{fill}" opacity="0.86"/>')
+        for frame in boundary_frames:
+            x = x_for(frame)
+            parts.append(f'<line x1="{x:.2f}" y1="{y + 4}" x2="{x:.2f}" y2="{y + 36}" stroke="#d8e887" stroke-width="1.2" opacity="0.70"/>')
+    parts.extend(
+        [
+            f'<text x="{left}" y="{height - 42}" fill="#8ee06a" font-family="Inter, Arial" font-size="12">green = exact/correct prediction</text>',
+            f'<text x="{left + 220}" y="{height - 42}" fill="#e46b5f" font-family="Inter, Arial" font-size="12">red = mismatch</text>',
+            f'<text x="{left + 390}" y="{height - 42}" fill="#d8e887" font-family="Inter, Arial" font-size="12">vertical lines = real transition boundaries</text>',
+            "</svg>",
+        ]
+    )
+    write_text(path, "\n".join(parts))
+
+
+def write_timeline_report(rows: list[dict], path: Path) -> None:
+    observed = [r for r in rows if r.get("status") == "observed_prediction"]
+    lines = [
+        "# Timeline Prediction Overlay Report",
+        "",
+        "This report aligns existing prediction CSV files to the exported episode timeline. It does not rerun training.",
+        "",
+        "## Task-Level Correctness",
+        "",
+    ]
+    for task in ["timeline_action", "timeline_subtask", "transition_detection", "next_action", "contact_prediction", "object_relevance"]:
+        task_rows = [r for r in observed if r["task"] == task]
+        if not task_rows:
+            lines.append(f"- {TASK_DISPLAY[task]}: no prediction rows found")
+            continue
+        correct = sum(int(r["correct"]) for r in task_rows)
+        lines.append(f"- {TASK_DISPLAY[task]}: {correct}/{len(task_rows)} correct ({correct / len(task_rows):.4f})")
+    lines.extend(
+        [
+            "",
+            "## Files",
+            "",
+            "- `timeline_overlay.csv`: prediction rows with frame positions.",
+            "- `timeline_overlay.svg`: visual overlay across the episode.",
+        ]
+    )
+    write_text(path, "\n".join(lines) + "\n")
+
+
+def run_alignment_stress(
+    X: np.ndarray,
+    manifest: list[dict],
+    windows: list[dict],
+    out_dir: Path,
+    args: argparse.Namespace,
+) -> list[dict]:
+    groups = modality_groups(manifest)
+    stress_groups = {
+        "motion_capture": groups["motion_capture"],
+        "pose_slam": groups["pose_slam"],
+        "inertial": groups["inertial"],
+        "language": groups["language"],
+        "motion_pose_inertial": np.unique(np.concatenate([groups["motion_capture"], groups["pose_slam"], groups["inertial"]])),
+    }
+    target_idx = block_indices(manifest, ["depth_confidence", "video_"])
+    n = X.shape[0]
+    train_idx, test_idx = chronological_split(n, args.test_fraction)
+    shifts = [-40, -20, -10, -5, 0, 5, 10, 20, 40]
+    rows: list[dict] = []
+    stride = int(windows[1]["start_frame"]) - int(windows[0]["start_frame"]) if len(windows) > 1 else 1
+    for group, q_idx in stress_groups.items():
+        q_train, q_test_all = standardize(X[train_idx][:, q_idx], X[test_idx][:, q_idx])
+        t_train, t_test_all = standardize(X[train_idx][:, target_idx], X[test_idx][:, target_idx])
+        projector_pred_all = ridge_predict(q_train, t_train, q_test_all, args.ridge_l2)
+        for shift in shifts:
+            valid = []
+            for local_i in range(len(test_idx)):
+                shifted_local = local_i + shift
+                if 0 <= shifted_local < len(test_idx):
+                    valid.append((local_i, shifted_local))
+            if not valid:
+                continue
+            original = np.asarray([a for a, _ in valid], dtype=np.int64)
+            shifted = np.asarray([b for _, b in valid], dtype=np.int64)
+            pred = projector_pred_all[shifted]
+            target = t_test_all[original]
+            metrics = retrieval_metrics(pred, target)
+            row = {
+                "query_group": group,
+                "query_display": GROUP_DISPLAY[group],
+                "target_group": "depth_plus_video",
+                "shift_windows": shift,
+                "shift_frames": int(shift * stride),
+                "status": "derived_perturbation",
+                **metrics,
+            }
+            rows.append(row)
+    write_csv(out_dir / "alignment_stress/alignment_shift_metrics.csv", rows)
+    render_alignment_svg(rows, out_dir / "alignment_stress/alignment_shift_curves.svg")
+    write_json(
+        out_dir / "alignment_stress/alignment_stress_summary.json",
+        {
+            "description": "Real feature windows are deliberately time-shifted at evaluation time to test cross-modal alignment sensitivity.",
+            "target_group": "depth_confidence + video_*",
+            "status_meaning": "derived_perturbation means the features are real but the time shift is an explicit diagnostic perturbation.",
+            "num_rows": len(rows),
+        },
+    )
+    write_alignment_report(rows, out_dir / "alignment_stress/ALIGNMENT_STRESS_REPORT.md")
+    return rows
+
+
+def render_alignment_svg(rows: list[dict], path: Path) -> None:
+    groups = sorted({r["query_group"] for r in rows})
+    width, height = 1020, 520
+    left, top, plot_w, plot_h = 94, 80, 810, 330
+    shifts = sorted({int(r["shift_windows"]) for r in rows})
+    if not shifts:
+        write_text(path, "<svg xmlns=\"http://www.w3.org/2000/svg\"></svg>")
+        return
+    min_shift, max_shift = min(shifts), max(shifts)
+    max_mrr = max(float(r["mrr"]) for r in rows) if rows else 1.0
+    max_mrr = max(max_mrr, 1e-6)
+    palette = ["#8ee06a", "#d8e887", "#7bd3ff", "#f0a45e", "#cba8ff"]
+
+    def x_for(shift: int) -> float:
+        return left + (shift - min_shift) / max(1, max_shift - min_shift) * plot_w
+
+    def y_for(mrr: float) -> float:
+        return top + plot_h - mrr / max_mrr * plot_h
+
+    parts = [
+        f'<svg xmlns="http://www.w3.org/2000/svg" width="{width}" height="{height}" viewBox="0 0 {width} {height}">',
+        '<rect width="100%" height="100%" fill="#10160f"/>',
+        '<text x="28" y="38" fill="#eef5e8" font-family="Inter, Arial" font-size="24" font-weight="700">Cross-Modal Alignment Stress Test</text>',
+        '<text x="28" y="64" fill="#a7b5a3" font-family="Inter, Arial" font-size="13">Query features are shifted in time; the target visual window remains the original held-out window.</text>',
+        f'<rect x="{left}" y="{top}" width="{plot_w}" height="{plot_h}" fill="#151d14" stroke="#34402f"/>',
+    ]
+    for tick in shifts:
+        x = x_for(tick)
+        parts.append(f'<line x1="{x:.2f}" y1="{top}" x2="{x:.2f}" y2="{top + plot_h}" stroke="#263024" stroke-width="1"/>')
+        parts.append(f'<text x="{x:.2f}" y="{top + plot_h + 24}" fill="#a7b5a3" font-family="Inter, Arial" font-size="11" text-anchor="middle">{tick}</text>')
+    parts.append(f'<text x="{left + plot_w / 2}" y="{height - 48}" fill="#cbd8c8" font-family="Inter, Arial" font-size="13" text-anchor="middle">shift in windows</text>')
+    parts.append(f'<text x="30" y="{top + plot_h / 2}" fill="#cbd8c8" font-family="Inter, Arial" font-size="13" transform="rotate(-90 30 {top + plot_h / 2})">MRR</text>')
+
+    for gi, group in enumerate(groups):
+        color = palette[gi % len(palette)]
+        group_rows = sorted([r for r in rows if r["query_group"] == group], key=lambda r: int(r["shift_windows"]))
+        points = [(x_for(int(r["shift_windows"])), y_for(float(r["mrr"]))) for r in group_rows]
+        if points:
+            d = " ".join(f"{x:.2f},{y:.2f}" for x, y in points)
+            parts.append(f'<polyline points="{d}" fill="none" stroke="{color}" stroke-width="2.4"/>')
+            for x, y in points:
+                parts.append(f'<circle cx="{x:.2f}" cy="{y:.2f}" r="4" fill="{color}"/>')
+            parts.append(f'<rect x="{left + plot_w + 28}" y="{top + gi * 25}" width="12" height="12" fill="{color}"/>')
+            parts.append(f'<text x="{left + plot_w + 46}" y="{top + 11 + gi * 25}" fill="#e7efe2" font-family="Inter, Arial" font-size="12">{html.escape(GROUP_DISPLAY.get(group, group))}</text>')
+    parts.append("</svg>")
+    write_text(path, "\n".join(parts))
+
+
+def write_alignment_report(rows: list[dict], path: Path) -> None:
+    lines = [
+        "# Cross-Modal Alignment Stress Report",
+        "",
+        "This diagnostic uses real held-out feature windows, then deliberately shifts the query modality in time at evaluation. The perturbation is derived; it is not treated as observed data.",
+        "",
+        "## Zero-Shift Versus Worst Shift",
+        "",
+    ]
+    for group in sorted({r["query_group"] for r in rows}):
+        group_rows = [r for r in rows if r["query_group"] == group]
+        zero = next((r for r in group_rows if int(r["shift_windows"]) == 0), None)
+        worst = min(group_rows, key=lambda r: float(r["mrr"])) if group_rows else None
+        if zero and worst:
+            lines.append(
+                f"- {GROUP_DISPLAY.get(group, group)}: zero-shift MRR={float(zero['mrr']):.4f}; "
+                f"worst shift={worst['shift_windows']} windows, MRR={float(worst['mrr']):.4f}"
+            )
+    lines.extend(
+        [
+            "",
+            "## Files",
+            "",
+            "- `alignment_shift_metrics.csv`: MRR/rank metrics for each query group and time shift.",
+            "- `alignment_shift_curves.svg`: MRR curves across time shifts.",
+            "- `alignment_stress_summary.json`: perturbation definition and status.",
+        ]
+    )
+    write_text(path, "\n".join(lines) + "\n")
+
+
+def build_provenance(
+    suite_dir: Path,
+    out_dir: Path,
+    X: np.ndarray,
+    starts: np.ndarray,
+    ends: np.ndarray,
+    manifest: list[dict],
+    summary: dict,
+    annotation: Path | None = None,
+) -> dict:
+    repo_root = suite_dir.parent.parent.resolve()
+    input_files = [
+        suite_dir / "shared_windows.npz",
+        suite_dir / "windows.csv",
+        suite_dir / "feature_manifest.json",
+        suite_dir / "summary_report.json",
+        suite_dir / "transition_detection/true_boundaries.csv",
+    ]
+    for task in ["timeline_action", "timeline_subtask", "transition_detection", "next_action", "contact_prediction", "object_relevance"]:
+        pred_path = suite_dir / task / "predictions.csv"
+        if pred_path.exists():
+            input_files.append(pred_path)
+    if annotation is not None and annotation.exists():
+        input_files.append(annotation)
+    provenance = {
+        "artifact_policy": "Only existing local artifacts are consumed. Missing labels/tasks are marked not_computed instead of filled.",
+        "source_suite_dir": public_artifact_path(suite_dir, repo_root),
+        "output_dir": public_artifact_path(out_dir, repo_root),
+        "shared_windows_shape": [int(X.shape[0]), int(X.shape[1])],
+        "starts_first_last": [int(starts[0]), int(starts[-1])],
+        "ends_first_last": [int(ends[0]), int(ends[-1])],
+        "feature_blocks": manifest,
+        "summary_report_core": {
+            "num_windows": summary.get("num_windows"),
+            "feature_dim": summary.get("feature_dim"),
+            "window_frames": summary.get("window_frames"),
+            "stride_frames": summary.get("stride_frames"),
+            "annotation": public_source_reference(summary.get("annotation")),
+        },
+        "input_file_hashes": {
+            public_artifact_path(path, repo_root): sha256(path)
+            for path in input_files
+            if path.exists()
+        },
+    }
+    write_json(out_dir / "provenance.json", provenance)
+    return provenance
+
+
+def write_index(out_dir: Path, provenance: dict, ablation_rows: list[dict], timeline_rows: list[dict], stress_rows: list[dict]) -> None:
+    lines = [
+        "# Single-Episode Diagnostics Index",
+        "",
+        "These outputs are local diagnostics built from the existing one-episode Xperience-10M artifacts. They are designed for manual verification while waiting for full multi-episode data access.",
+        "",
+        "## Generated Analyses",
+        "",
+        "- `modality_ablation/`: compact ridge-head ablations across real feature blocks.",
+        "- `timeline_overlay/`: existing prediction CSVs aligned to the episode timeline.",
+        "- `alignment_stress/`: cross-modal retrieval under explicit time-shift perturbations.",
+        "- `provenance.json`: input hashes, feature dimensions, and source artifact identifiers.",
+        "",
+        "## Validity Boundaries",
+        "",
+        "- This is a single-episode diagnostic, not a full Xperience-10M benchmark.",
+        "- Rows marked `not_computed` are intentionally left blank when train labels or valid splits are unavailable.",
+        "- Rows marked `derived_perturbation` use real features with deliberate time shifts for stress testing.",
+        "",
+        "## Counts",
+        "",
+        f"- Ablation rows: {len(ablation_rows)}; computed: {sum(1 for r in ablation_rows if r.get('status') == 'computed')}.",
+        f"- Timeline overlay rows: {sum(1 for r in timeline_rows if r.get('status') == 'observed_prediction')}.",
+        f"- Alignment stress rows: {len(stress_rows)}.",
+        f"- Shared feature shape: {provenance['shared_windows_shape'][0]} windows x {provenance['shared_windows_shape'][1]} features.",
+    ]
+    write_text(out_dir / "README.md", "\n".join(lines) + "\n")
+
+
+def main() -> None:
+    args = parse_args()
+    suite_dir = args.suite_dir.resolve()
+    out_dir = args.output_dir.resolve()
+    X, starts, ends, windows, manifest, summary = load_inputs(suite_dir)
+    object_targets = None
+    if args.annotation is not None:
+        object_targets = load_object_targets_from_annotation(args.annotation, windows, args.homie_toolkit)
+        write_csv(
+            out_dir / "object_labels/window_object_labels.csv",
+            object_targets["rows"],
+            ["window_index", "start_frame", "end_frame", "center_frame", "objects", "object_count"],
+        )
+        write_json(
+            out_dir / "object_labels/object_vocab.json",
+            {
+                "vocab": object_targets["vocab"],
+                "num_objects": len(object_targets["vocab"]),
+                "source_annotation": object_targets["annotation"],
+                "source_toolkit": object_targets["toolkit"],
+                "source_note": object_targets["source_note"],
+            },
+        )
+    provenance = build_provenance(suite_dir, out_dir, X, starts, ends, manifest, summary, args.annotation)
+    ablation_rows = run_modality_ablation(X, windows, manifest, suite_dir, out_dir, args, object_targets)
+    timeline_rows = run_timeline_overlay(suite_dir, windows, out_dir)
+    stress_rows = run_alignment_stress(X, manifest, windows, out_dir, args)
+    write_index(out_dir, provenance, ablation_rows, timeline_rows, stress_rows)
+    print(f"Wrote diagnostics to {out_dir}")
+    print(f"Ablation computed rows: {sum(1 for r in ablation_rows if r.get('status') == 'computed')}/{len(ablation_rows)}")
+    print(f"Timeline observed rows: {sum(1 for r in timeline_rows if r.get('status') == 'observed_prediction')}")
+    print(f"Alignment stress rows: {len(stress_rows)}")
+
+
+if __name__ == "__main__":
+    main()

scripts/validate_mirror_parity.py

@@ -41,6 +41,7 @@ DATA_FILES = [
     "research_direction_extensions.json",
     "research_directions.json",
     "scope_claims_audit.json",
+    "single_episode_explorer.json",
     "source_alignment_audit.json",
     "summary_metrics.json",
     "task_surface_integrity.json",
@@ -77,7 +78,9 @@ SCRIPT_FILES = [
     "build_quality_gates.py",
     "build_public_surface_qa.py",
     "build_rendered_site_check.py",
+    "build_single_episode_explorer.py",
     "build_research_takeaways.py",
+    "single_episode_diagnostics.py",
     "verify_live_publication.py",
     "validate_mirror_parity.py",
     "validate_publication_package.py",
@@ -92,9 +95,22 @@ WEBSITE_FILES = [
     "apple-touch-icon.png",
     "favicon.png",
     "index.html",
+    "single_episode_explorer.html",
     "site.webmanifest",
 ]
 
+RESULT_FILES = [
+    "single_episode_diagnostics/provenance.json",
+    "single_episode_diagnostics/README.md",
+    "single_episode_diagnostics/modality_ablation/ablation_metrics.csv",
+    "single_episode_diagnostics/modality_ablation/ablation_summary.json",
+    "single_episode_diagnostics/object_labels/object_vocab.json",
+    "single_episode_diagnostics/object_labels/window_object_labels.csv",
+    "single_episode_diagnostics/timeline_overlay/timeline_overlay.csv",
+    "single_episode_diagnostics/alignment_stress/alignment_shift_metrics.csv",
+    "single_episode_diagnostics/alignment_stress/alignment_stress_summary.json",
+]
+
 DOC_FILES = [
     "QUALITY_GATES.md",
     "EVALUATION_PROTOCOL.md",
@@ -236,6 +252,20 @@ def build_report(hf_root: Path) -> dict:
             )
         )
 
+    for filename in RESULT_FILES:
+        groups.append(
+            parity_group(
+                f"results/{filename}",
+                ROOT / "results" / filename,
+                {
+                    "hf_space": hf_root / "space/results" / filename,
+                    "hf_artifacts": hf_root / "artifacts/results" / filename,
+                    "hf_model": hf_root / "model/results" / filename,
+                },
+                hf_root,
+            )
+        )
+
     for filename in DOC_FILES:
         groups.append(
             parity_group(
@@ -293,6 +323,12 @@ def build_report(hf_root: Path) -> dict:
                 if not any(failure["group"].startswith("website/") for failure in failures)
                 else "fail",
             },
+            {
+                "name": "repo_hf_diagnostic_result_parity",
+                "status": "pass"
+                if not any(failure["group"].startswith("results/") for failure in failures)
+                else "fail",
+            },
             {
                 "name": "repo_hf_quality_doc_parity",
                 "status": "pass"