| { |
| "generated_at_utc": "2026-06-22T11:00:00+00:00", |
| "immediate_actions": [ |
| { |
| "artifact": "docs/data/task_method_20_gap_audit.json", |
| "id": "gap_audit", |
| "purpose": "Verify the 180/180 scored result records and keep proxy flags reproducible." |
| }, |
| { |
| "artifact": "scripts/omni/score_model_output_probes.py", |
| "id": "model_output_probe", |
| "purpose": "Rescore verified model-output probes when new held-out artifacts arrive without fabricating unsupported cells." |
| }, |
| { |
| "artifact": "scripts/omni/launch_all_task_model_scoring_when_free.sh", |
| "id": "guarded_gpu_launcher", |
| "purpose": "Launch future replacement scoring runs only after enough private GPU capacity is idle." |
| } |
| ], |
| "methods": { |
| "cosmos3_nano_future_window": { |
| "kind": "partial_128_episode_world_model_overlay", |
| "label": "Cosmos3-Nano Future Window", |
| "proxy_scored_task_count": 0, |
| "result_record_count": 20, |
| "scope": "128 selected episodes, held-out test", |
| "scored_task_count": 20, |
| "scoreless_task_count": 0, |
| "status_counts": { |
| "scored": 20 |
| } |
| }, |
| "cosmos3_super_reasoner": { |
| "kind": "partial_128_episode_foundation_model_overlay", |
| "label": "Cosmos3-Super Reasoner", |
| "proxy_scored_task_count": 0, |
| "result_record_count": 20, |
| "scope": "128 selected episodes, held-out test", |
| "scored_task_count": 20, |
| "scoreless_task_count": 0, |
| "status_counts": { |
| "scored": 20 |
| } |
| }, |
| "metadata128_neural_mlp": { |
| "kind": "partial_128_episode_aligned_baseline", |
| "label": "128ep Aligned NN", |
| "proxy_scored_task_count": 1, |
| "result_record_count": 20, |
| "scope": "128 selected episodes, JSONL metadata/text plus staged sensor-block targets where available", |
| "scored_task_count": 20, |
| "scoreless_task_count": 0, |
| "status_counts": { |
| "proxy_scored": 1, |
| "scored": 19 |
| } |
| }, |
| "metadata128_simple": { |
| "kind": "partial_128_episode_aligned_baseline", |
| "label": "128ep Aligned Simple", |
| "proxy_scored_task_count": 1, |
| "result_record_count": 20, |
| "scope": "128 selected episodes, JSONL metadata/text plus staged sensor-block targets where available", |
| "scored_task_count": 20, |
| "scoreless_task_count": 0, |
| "status_counts": { |
| "proxy_scored": 1, |
| "scored": 19 |
| } |
| }, |
| "minimal": { |
| "kind": "full_20_task_baseline", |
| "label": "Minimal", |
| "proxy_scored_task_count": 0, |
| "result_record_count": 20, |
| "scope": "1 public sample episode", |
| "scored_task_count": 20, |
| "scoreless_task_count": 0, |
| "status_counts": { |
| "scored": 20 |
| } |
| }, |
| "neural_mlp": { |
| "kind": "full_20_task_baseline", |
| "label": "Neural MLP", |
| "proxy_scored_task_count": 0, |
| "result_record_count": 20, |
| "scope": "1 public sample episode", |
| "scored_task_count": 20, |
| "scoreless_task_count": 0, |
| "status_counts": { |
| "scored": 20 |
| } |
| }, |
| "qwen3_omni_v6_lora": { |
| "kind": "partial_128_episode_foundation_model_overlay", |
| "label": "Qwen3-Omni v6 LoRA", |
| "proxy_scored_task_count": 0, |
| "result_record_count": 20, |
| "scope": "128 selected episodes, held-out test", |
| "scored_task_count": 20, |
| "scoreless_task_count": 0, |
| "status_counts": { |
| "scored": 20 |
| } |
| }, |
| "raw128_neural_mlp": { |
| "kind": "complete_128_episode_raw_feature_baseline", |
| "label": "128ep Raw NN", |
| "proxy_scored_task_count": 2, |
| "result_record_count": 20, |
| "scope": "128 selected episodes, staged 4430-dim sensor NPZ features; 2 compact proxy axes", |
| "scored_task_count": 20, |
| "scoreless_task_count": 0, |
| "status_counts": { |
| "proxy_scored": 2, |
| "scored": 18 |
| } |
| }, |
| "raw128_simple": { |
| "kind": "complete_128_episode_raw_feature_baseline", |
| "label": "128ep Raw Simple", |
| "proxy_scored_task_count": 2, |
| "result_record_count": 20, |
| "scope": "128 selected episodes, staged 4430-dim sensor NPZ features; 2 compact proxy axes", |
| "scored_task_count": 20, |
| "scoreless_task_count": 0, |
| "status_counts": { |
| "proxy_scored": 2, |
| "scored": 18 |
| } |
| } |
| }, |
| "missing_by_method": {}, |
| "missing_by_status": {}, |
| "missing_by_task": {}, |
| "missing_records": [], |
| "proxy_records": [ |
| { |
| "method": "128ep Raw Simple", |
| "metric_key": "macro_f1", |
| "reason": "documented compact proxy completion for this raw128 task axis", |
| "series_id": "raw128_simple", |
| "source": "results/omni_finetune/a100_128_raw20_task_baselines_complete20_proxy_20260616T091500Z/simple_raw128/interaction_text_prediction/metrics.json", |
| "task_id": "interaction_text_prediction", |
| "task_label": "Interaction Text Prediction", |
| "task_number": 15 |
| }, |
| { |
| "method": "128ep Raw NN", |
| "metric_key": "macro_f1", |
| "reason": "documented compact proxy completion for this raw128 task axis", |
| "series_id": "raw128_neural_mlp", |
| "source": "results/omni_finetune/a100_128_raw20_task_baselines_complete20_proxy_20260616T091500Z/neural_mlp_raw128/interaction_text_prediction/metrics.json", |
| "task_id": "interaction_text_prediction", |
| "task_label": "Interaction Text Prediction", |
| "task_number": 15 |
| }, |
| { |
| "method": "128ep Aligned Simple", |
| "metric_key": "mrr", |
| "reason": "paired camera-view embeddings are absent from the 128 JSONL/feature export; metadata features retrieve the synchronized same-window depth/audio block as a documented compact synchronization proxy", |
| "series_id": "metadata128_simple", |
| "source": "results/omni_finetune/a100_128_metadata_task_baselines_20260616_v2/camera_view_sync_retrieval/metrics.json", |
| "task_id": "camera_view_sync_retrieval", |
| "task_label": "Camera-View Synchronization Retrieval", |
| "task_number": 19 |
| }, |
| { |
| "method": "128ep Aligned NN", |
| "metric_key": "mrr", |
| "reason": "paired camera-view embeddings are absent from the 128 JSONL/feature export; metadata features retrieve the synchronized same-window depth/audio block as a documented compact synchronization proxy", |
| "series_id": "metadata128_neural_mlp", |
| "source": "results/omni_finetune/a100_128_metadata_task_baselines_20260616_v2/neural_mlp/camera_view_sync_retrieval/metrics.json", |
| "task_id": "camera_view_sync_retrieval", |
| "task_label": "Camera-View Synchronization Retrieval", |
| "task_number": 19 |
| }, |
| { |
| "method": "128ep Raw Simple", |
| "metric_key": "mrr", |
| "reason": "documented compact proxy completion for this raw128 task axis", |
| "series_id": "raw128_simple", |
| "source": "results/omni_finetune/a100_128_raw20_task_baselines_complete20_proxy_20260616T091500Z/simple_raw128/camera_view_sync_retrieval/metrics.json", |
| "task_id": "camera_view_sync_retrieval", |
| "task_label": "Camera-View Synchronization Retrieval", |
| "task_number": 19 |
| }, |
| { |
| "method": "128ep Raw NN", |
| "metric_key": "mrr", |
| "reason": "documented compact proxy completion for this raw128 task axis", |
| "series_id": "raw128_neural_mlp", |
| "source": "results/omni_finetune/a100_128_raw20_task_baselines_complete20_proxy_20260616T091500Z/neural_mlp_raw128/camera_view_sync_retrieval/metrics.json", |
| "task_id": "camera_view_sync_retrieval", |
| "task_label": "Camera-View Synchronization Retrieval", |
| "task_number": 19 |
| } |
| ], |
| "score_summary": { |
| "method_count": 9, |
| "method_task_record_count": 180, |
| "proxy_scored_method_task_count": 6, |
| "scored_method_task_count": 180, |
| "scoreless_method_task_count": 0, |
| "task_count": 20 |
| }, |
| "source_matrix": "docs/data/task_method_20_result_matrix.json", |
| "status": "pass", |
| "target_policy": { |
| "numeric_score_gate": "A method-task cell is numeric only when a runner or verified package emits that exact task target and metric.", |
| "proxy_policy": "Proxy scores are allowed only when the matrix marks them as proxy_scored and keeps the reason/source attached.", |
| "scoreless_cell_policy": "If future unsupported or not-evaluated cells appear, they must stay explicit in the public matrix instead of being hidden or backfilled with proxy model numbers. The current release has zero scoreless cells." |
| }, |
| "title": "Task Method 20-Result Completion Audit" |
| } |
|
|
