Datasets:
File size: 3,361 Bytes
45deef7 db94f6b 45deef7 db94f6b 744bfa2 45deef7 744bfa2 45deef7 9c70200 5618ed1 db94f6b 45deef7 744bfa2 45deef7 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 | # Task Method 20-Result Gap Audit
Generated: `2026-06-20T18:44:28+00:00`
This audit is the explicit gap ledger for the 9-method x 20-task result matrix.
It keeps missing cells visible while preserving the rule that a numeric score
requires a real task target and source artifact.
## Score Summary
- Method-task records: `180`
- Numeric scored records: `180`
- Scoreless records: `0`
- Proxy-scored records: `6`
- Source matrix: [`docs/data/task_method_20_result_matrix.json`](docs/data/task_method_20_result_matrix.json)
## Method Coverage
| Method | ID | Scored | Scoreless | Proxy | Status counts |
| --- | --- | --- | --- | --- | --- |
| Minimal | minimal | 20/20 | 0 | 0 | scored: 20 |
| Neural MLP | neural_mlp | 20/20 | 0 | 0 | scored: 20 |
| 128ep Aligned Simple | metadata128_simple | 20/20 | 0 | 1 | proxy_scored: 1, scored: 19 |
| 128ep Aligned NN | metadata128_neural_mlp | 20/20 | 0 | 1 | proxy_scored: 1, scored: 19 |
| 128ep Raw Simple | raw128_simple | 20/20 | 0 | 2 | proxy_scored: 2, scored: 18 |
| 128ep Raw NN | raw128_neural_mlp | 20/20 | 0 | 2 | proxy_scored: 2, scored: 18 |
| Qwen3-Omni v6 LoRA | qwen3_omni_v6_lora | 20/20 | 0 | 0 | scored: 20 |
| Cosmos3-Super Reasoner | cosmos3_super_reasoner | 20/20 | 0 | 0 | scored: 20 |
| Cosmos3-Nano Future Window | cosmos3_nano_future_window | 20/20 | 0 | 0 | scored: 20 |
## Gap Classes
| Status | Count | Next step |
| --- | --- | --- |
## Scoreless Records
| Task | Task label | Method | Status | Required evidence |
| --- | --- | --- | --- | --- |
## Proxy Records
| Task | Task label | Method | Metric | Proxy note |
| --- | --- | --- | --- | --- |
| 15 | Interaction Text Prediction | 128ep Raw Simple | macro_f1 | documented compact proxy completion for this raw128 task axis |
| 15 | Interaction Text Prediction | 128ep Raw NN | macro_f1 | documented compact proxy completion for this raw128 task axis |
| 19 | Camera-View Synchronization Retrieval | 128ep Aligned Simple | mrr | 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 |
| 19 | Camera-View Synchronization Retrieval | 128ep Aligned NN | mrr | 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 |
| 19 | Camera-View Synchronization Retrieval | 128ep Raw Simple | mrr | documented compact proxy completion for this raw128 task axis |
| 19 | Camera-View Synchronization Retrieval | 128ep Raw NN | mrr | documented compact proxy completion for this raw128 task axis |
## Immediate Actions
- Keep [`docs/data/task_method_20_gap_audit.json`](docs/data/task_method_20_gap_audit.json) next to the radar and matrix so readers can distinguish scored, proxy-scored, and scoreless cells.
- Use [`scripts/omni/score_model_output_probes.py`](scripts/omni/score_model_output_probes.py) to check whether train/validation/test model outputs are present before trying to extend Qwen3/Cosmos to all 20 task contracts.
- Use [`scripts/omni/launch_all_task_model_scoring_when_free.sh`](scripts/omni/launch_all_task_model_scoring_when_free.sh) as the guarded waiter for a real all-task scoring command when private GPU capacity is available.
|