TASK_METHOD_20_RESULT_MATRIX.md

# Task Method 20-Result Matrix

Every method has one record for each of the 20 unified task contracts. Numeric scores appear only where a committed runner or verified package produced that task target.

Legend: `score` = direct numeric task score and `proxy` = documented compact substitute target. The current public matrix is complete at 180/180 scored records; unsupported/not-evaluated labels are retained only for future regression audits.

| Method | Records | Scored | Proxy scored | Scoreless | Status counts |
| --- | ---: | ---: | ---: | ---: | --- |
| Minimal | 20 | 20 | 0 | 0 | scored 20 |
| Neural MLP | 20 | 20 | 0 | 0 | scored 20 |
| 128ep Aligned Simple | 20 | 20 | 1 | 0 | proxy scored 1, scored 19 |
| 128ep Aligned NN | 20 | 20 | 1 | 0 | proxy scored 1, scored 19 |
| 128ep Raw Simple | 20 | 20 | 2 | 0 | proxy scored 2, scored 18 |
| 128ep Raw NN | 20 | 20 | 2 | 0 | proxy scored 2, scored 18 |
| Qwen3-Omni v6 LoRA | 20 | 20 | 0 | 0 | scored 20 |
| Cosmos3-Super Reasoner | 20 | 20 | 0 | 0 | scored 20 |
| Cosmos3-Nano Future Window | 20 | 20 | 0 | 0 | scored 20 |

## Compact Score Matrix

Cells show `raw metric value`, then `direct/proxy; normalized radar value; metric key`. The raw metric is the value to cite; the normalized value is the exact linear 0-1 score retained in JSON. The SVG radar uses sqrt(normalized score) only for visual radius, so low but real differences remain visible without changing the table values.

| # | Task | Min | NN | 128-S | 128-NN | 128-RS | 128-RN | Qwen3 | C3-S | C3-N |
| ---: | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| 01 | Action Recognition | 0.0500<br><sub>direct; norm 0.050; macro_f1</sub> | 0.0148<br><sub>direct; norm 0.015; macro_f1</sub> | 0.0083<br><sub>direct; norm 0.008; macro_f1</sub> | 0.0042<br><sub>direct; norm 0.004; macro_f1</sub> | 0.0029<br><sub>direct; norm 0.003; macro_f1</sub> | 0.0015<br><sub>direct; norm 0.001; macro_f1</sub> | 0.0029<br><sub>direct; norm 0.003; action_macro_f1</sub> | 0.0008<br><sub>direct; norm 0.001; action_macro_f1</sub> | 0.0079<br><sub>direct; norm 0.008; action_accuracy_from_retrieved_future</sub> |
| 02 | Procedure Step Recognition | 0.0506<br><sub>direct; norm 0.051; macro_f1</sub> | 0.0281<br><sub>direct; norm 0.028; macro_f1</sub> | 0.0002<br><sub>direct; norm 0.000; macro_f1</sub> | 0.0001<br><sub>direct; norm 0.000; macro_f1</sub> | 0.0000<br><sub>direct; norm 0.000; macro_f1</sub> | 0.0001<br><sub>direct; norm 0.000; macro_f1</sub> | 0.0037<br><sub>direct; norm 0.004; subtask_accuracy</sub> | 0.0000<br><sub>direct; norm 0.000; subtask_accuracy</sub> | 0.0000<br><sub>direct; norm 0.000; timeline_subtask_macro_f1</sub> |
| 03 | Action Boundary Detection | 0.6118<br><sub>direct; norm 0.612; macro_f1</sub> | 0.5862<br><sub>direct; norm 0.586; macro_f1</sub> | 0.2965<br><sub>direct; norm 0.297; macro_f1</sub> | 0.4842<br><sub>direct; norm 0.484; macro_f1</sub> | 0.4204<br><sub>direct; norm 0.420; macro_f1</sub> | 0.4902<br><sub>direct; norm 0.490; macro_f1</sub> | 0.9898<br><sub>direct; norm 0.990; transition_accuracy</sub> | 0.3683<br><sub>direct; norm 0.368; transition_accuracy</sub> | 0.9683<br><sub>direct; norm 0.968; transition_accuracy</sub> |
| 04 | Next-Action Prediction | 0.0593<br><sub>direct; norm 0.059; macro_f1</sub> | 0.0419<br><sub>direct; norm 0.042; macro_f1</sub> | 0.0065<br><sub>direct; norm 0.007; macro_f1</sub> | 0.0049<br><sub>direct; norm 0.005; macro_f1</sub> | 0.0033<br><sub>direct; norm 0.003; macro_f1</sub> | 0.0018<br><sub>direct; norm 0.002; macro_f1</sub> | 0.0431<br><sub>direct; norm 0.043; next_action_accuracy</sub> | 0.0134<br><sub>direct; norm 0.013; next_action_accuracy</sub> | 0.0079<br><sub>direct; norm 0.008; action_accuracy_from_retrieved_future</sub> |
| 05 | Hand Trajectory Forecasting | 0.8647<br><sub>direct; norm 0.125; mpjpe</sub> | 0.1079<br><sub>direct; norm 1.000; mpjpe</sub> | 8.817<br><sub>direct; norm 0.012; mpjpe</sub> | 0.4294<br><sub>direct; norm 0.251; mpjpe</sub> | 0.2729<br><sub>direct; norm 0.395; mae</sub> | 0.1848<br><sub>direct; norm 0.584; mae</sub> | 0.7216<br><sub>direct; norm 0.149; hand_trajectory_forecast_mrr</sub> | 0.8915<br><sub>direct; norm 0.121; hand_trajectory_forecast_mrr</sub> | 0.6913<br><sub>direct; norm 0.156; hand_trajectory_forecast_mrr</sub> |
| 06 | Contact State Prediction | 1.000<br><sub>direct; norm 1.000; macro_f1</sub> | 1.000<br><sub>direct; norm 1.000; macro_f1</sub> | 0.4381<br><sub>direct; norm 0.438; macro_f1</sub> | 0.5683<br><sub>direct; norm 0.568; macro_f1</sub> | 0.8870<br><sub>direct; norm 0.887; macro_f1</sub> | 1.000<br><sub>direct; norm 1.000; macro_f1</sub> | 0.8177<br><sub>direct; norm 0.818; contact_accuracy</sub> | 0.3214<br><sub>direct; norm 0.321; contact_accuracy</sub> | 0.7434<br><sub>direct; norm 0.743; contact_accuracy</sub> |
| 07 | Object Relevance Prediction | 0.1803<br><sub>direct; norm 0.180; micro_f1</sub> | 0.1679<br><sub>direct; norm 0.168; micro_f1</sub> | 0.1776<br><sub>direct; norm 0.178; micro_f1</sub> | 0.1866<br><sub>direct; norm 0.187; micro_f1</sub> | 0.0655<br><sub>direct; norm 0.066; micro_f1</sub> | 0.1766<br><sub>direct; norm 0.177; micro_f1</sub> | 0.3065<br><sub>direct; norm 0.306; object_micro_f1</sub> | 0.1370<br><sub>direct; norm 0.137; object_micro_f1</sub> | 0.0005<br><sub>direct; norm 0.000; object_relevance_micro_f1</sub> |
| 08 | Language Grounding | 0.0160<br><sub>direct; norm 0.016; mrr</sub> | 0.0168<br><sub>direct; norm 0.017; mrr</sub> | 0.0023<br><sub>direct; norm 0.002; mrr</sub> | 0.0082<br><sub>direct; norm 0.008; mrr</sub> | 0.0111<br><sub>direct; norm 0.011; mrr</sub> | 0.0063<br><sub>direct; norm 0.006; mrr</sub> | 0.8764<br><sub>direct; norm 0.876; caption_grounding_mrr</sub> | 0.3064<br><sub>direct; norm 0.306; caption_grounding_iou</sub> | 0.5221<br><sub>direct; norm 0.522; caption_grounding_mrr</sub> |
| 09 | Cross-Modal Retrieval | 0.2693<br><sub>direct; norm 0.269; mrr</sub> | 0.1300<br><sub>direct; norm 0.130; mrr</sub> | 0.0026<br><sub>direct; norm 0.003; mrr</sub> | 0.0026<br><sub>direct; norm 0.003; mrr</sub> | 0.0035<br><sub>direct; norm 0.003; mrr</sub> | 0.0025<br><sub>direct; norm 0.003; mrr</sub> | 0.5080<br><sub>direct; norm 0.508; cross_modal_retrieval_mrr</sub> | 0.6628<br><sub>direct; norm 0.663; cross_modal_retrieval_mrr</sub> | 0.0221<br><sub>direct; norm 0.022; future_retrieval_mrr</sub> |
| 10 | Cross-Modal Reconstruction | -0.0153<br><sub>direct; norm 0.000; r2</sub> | -0.0102<br><sub>direct; norm 0.000; r2</sub> | -190.66<br><sub>direct; norm 0.000; r2</sub> | -0.4348<br><sub>direct; norm 0.000; r2</sub> | -1.345<br><sub>direct; norm 0.000; r2</sub> | -1.397<br><sub>direct; norm 0.000; r2</sub> | 0.9671<br><sub>direct; norm 0.967; modality_reconstruction_mrr</sub> | 0.9939<br><sub>direct; norm 0.994; modality_reconstruction_mrr</sub> | 0.0003<br><sub>direct; norm 0.000; feature_reconstruction_quality</sub> |
| 11 | Temporal Order Verification | 0.5400<br><sub>direct; norm 0.540; f1</sub> | 0.8520<br><sub>direct; norm 0.852; f1</sub> | 0.4199<br><sub>direct; norm 0.420; f1</sub> | 0.8252<br><sub>direct; norm 0.825; f1</sub> | 0.4982<br><sub>direct; norm 0.498; macro_f1</sub> | 0.8030<br><sub>direct; norm 0.803; macro_f1</sub> | 0.4098<br><sub>direct; norm 0.410; temporal_order_f1</sub> | 0.6286<br><sub>direct; norm 0.629; temporal_order_f1</sub> | 0.5954<br><sub>direct; norm 0.595; temporal_order_f1</sub> |
| 12 | Multimodal Synchronization Detection | 0.5052<br><sub>direct; norm 0.505; f1</sub> | 0.7153<br><sub>direct; norm 0.715; f1</sub> | 0.4998<br><sub>direct; norm 0.500; f1</sub> | 0.7774<br><sub>direct; norm 0.777; f1</sub> | 0.4959<br><sub>direct; norm 0.496; macro_f1</sub> | 0.8273<br><sub>direct; norm 0.827; macro_f1</sub> | 0.3345<br><sub>direct; norm 0.334; misalignment_detection_f1</sub> | 0.3727<br><sub>direct; norm 0.373; misalignment_detection_f1</sub> | 0.4772<br><sub>direct; norm 0.477; misalignment_detection_f1</sub> |
| 13 | Long-Horizon Next-Action Forecasting | 0.0750<br><sub>direct; norm 0.075; macro_f1</sub> | 0.0655<br><sub>direct; norm 0.065; macro_f1</sub> | 0.0046<br><sub>direct; norm 0.005; macro_f1</sub> | 0.0030<br><sub>direct; norm 0.003; macro_f1</sub> | 0.0024<br><sub>direct; norm 0.002; macro_f1</sub> | 0.0011<br><sub>direct; norm 0.001; macro_f1</sub> | 0.0023<br><sub>direct; norm 0.002; long_horizon_next_action_macro_f1</sub> | 0.0088<br><sub>direct; norm 0.009; long_horizon_next_action_macro_f1</sub> | 0.0025<br><sub>direct; norm 0.002; long_horizon_next_action_macro_f1</sub> |
| 14 | Long-Horizon Next-Subtask Forecasting | 0.0455<br><sub>direct; norm 0.045; macro_f1</sub> | 0.0507<br><sub>direct; norm 0.051; macro_f1</sub> | 0.0001<br><sub>direct; norm 0.000; macro_f1</sub> | 0.0000<br><sub>direct; norm 0.000; macro_f1</sub> | 0.0000<br><sub>direct; norm 0.000; macro_f1</sub> | 0.0000<br><sub>direct; norm 0.000; macro_f1</sub> | 0.0042<br><sub>direct; norm 0.004; next_subtask_forecast_macro_f1</sub> | 0.0000<br><sub>direct; norm 0.000; next_subtask_forecast_macro_f1</sub> | 0.0066<br><sub>direct; norm 0.007; next_subtask_forecast_macro_f1</sub> |
| 15 | Interaction Text Prediction | 0.0444<br><sub>direct; norm 0.044; macro_f1</sub> | 0.0381<br><sub>direct; norm 0.038; macro_f1</sub> | 0.0000<br><sub>direct; norm 0.000; macro_f1</sub> | 0.0000<br><sub>direct; norm 0.000; macro_f1</sub> | 0.0126<br><sub>proxy; norm 0.013; macro_f1</sub> | 0.0098<br><sub>proxy; norm 0.010; macro_f1</sub> | 0.4319<br><sub>direct; norm 0.432; macro_f1</sub> | 0.1795<br><sub>direct; norm 0.179; macro_f1</sub> | 0.1788<br><sub>direct; norm 0.179; macro_f1</sub> |
| 16 | Action-Object Relation Prediction | 0.0000<br><sub>direct; norm 0.000; macro_f1</sub> | 0.0000<br><sub>direct; norm 0.000; macro_f1</sub> | 0.0000<br><sub>direct; norm 0.000; macro_f1</sub> | 0.0000<br><sub>direct; norm 0.000; macro_f1</sub> | 0.0000<br><sub>direct; norm 0.000; macro_f1</sub> | 0.0000<br><sub>direct; norm 0.000; macro_f1</sub> | 0.0002<br><sub>direct; norm 0.000; action_object_relation_macro_f1</sub> | 0.0000<br><sub>direct; norm 0.000; action_object_relation_macro_f1</sub> | 0.0028<br><sub>direct; norm 0.003; action_object_relation_macro_f1</sub> |
| 17 | Future Object-Set Forecasting | 0.1694<br><sub>direct; norm 0.169; micro_f1</sub> | 0.1972<br><sub>direct; norm 0.197; micro_f1</sub> | 0.1766<br><sub>direct; norm 0.177; micro_f1</sub> | 0.1742<br><sub>direct; norm 0.174; micro_f1</sub> | 0.0647<br><sub>direct; norm 0.065; micro_f1</sub> | 0.1752<br><sub>direct; norm 0.175; micro_f1</sub> | 0.1659<br><sub>direct; norm 0.166; object_set_forecast_micro_f1</sub> | 0.0009<br><sub>direct; norm 0.001; object_set_forecast_micro_f1</sub> | 0.0178<br><sub>direct; norm 0.018; object_set_forecast_micro_f1</sub> |
| 18 | IMU-to-Hand Pose Reconstruction | 0.0420<br><sub>direct; norm 1.000; mae</sub> | 0.0426<br><sub>direct; norm 0.988; mae</sub> | 0.2295<br><sub>direct; norm 0.183; mae</sub> | 0.2556<br><sub>direct; norm 0.165; mae</sub> | 0.2294<br><sub>direct; norm 0.183; mae</sub> | 0.2530<br><sub>direct; norm 0.166; mae</sub> | 0.9642<br><sub>direct; norm 0.044; imu_to_hand_pose_mrr</sub> | 0.9897<br><sub>direct; norm 0.042; imu_to_hand_pose_mrr</sub> | 0.9920<br><sub>direct; norm 0.042; imu_to_hand_pose_mrr</sub> |
| 19 | Camera-View Synchronization Retrieval | 0.4943<br><sub>direct; norm 0.494; mrr</sub> | 0.2409<br><sub>direct; norm 0.241; mrr</sub> | 0.0021<br><sub>proxy; norm 0.002; mrr</sub> | 0.0027<br><sub>proxy; norm 0.003; mrr</sub> | 0.0027<br><sub>proxy; norm 0.003; mrr</sub> | 0.0025<br><sub>proxy; norm 0.003; mrr</sub> | 0.6588<br><sub>direct; norm 0.659; camera_view_sync_retrieval_mrr</sub> | 0.9980<br><sub>direct; norm 0.998; camera_view_sync_retrieval_mrr</sub> | 0.9990<br><sub>direct; norm 0.999; camera_view_sync_retrieval_mrr</sub> |
| 20 | Time-to-Next-Transition Regression | 10.54<br><sub>direct; norm 1.000; mae</sub> | 10.55<br><sub>direct; norm 0.998; mae</sub> | 624.81<br><sub>direct; norm 0.017; mae</sub> | 41.47<br><sub>direct; norm 0.254; mae</sub> | 52.33<br><sub>direct; norm 0.201; mae</sub> | 42.37<br><sub>direct; norm 0.249; mae</sub> | 134.07<br><sub>direct; norm 0.079; time_to_transition_mae</sub> | 52.95<br><sub>direct; norm 0.199; time_to_transition_mae</sub> | 33.81<br><sub>direct; norm 0.312; time_to_transition_mae</sub> |

## Status Matrix

| # | Task | Min | NN | 128-S | 128-NN | 128-RS | 128-RN | Qwen3 | C3-S | C3-N |
| ---: | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| 01 | Action Recognition | score | score | score | score | score | score | score | score | score |
| 02 | Procedure Step Recognition | score | score | score | score | score | score | score | score | score |
| 03 | Action Boundary Detection | score | score | score | score | score | score | score | score | score |
| 04 | Next-Action Prediction | score | score | score | score | score | score | score | score | score |
| 05 | Hand Trajectory Forecasting | score | score | score | score | score | score | score | score | score |
| 06 | Contact State Prediction | score | score | score | score | score | score | score | score | score |
| 07 | Object Relevance Prediction | score | score | score | score | score | score | score | score | score |
| 08 | Language Grounding | score | score | score | score | score | score | score | score | score |
| 09 | Cross-Modal Retrieval | score | score | score | score | score | score | score | score | score |
| 10 | Cross-Modal Reconstruction | score | score | score | score | score | score | score | score | score |
| 11 | Temporal Order Verification | score | score | score | score | score | score | score | score | score |
| 12 | Multimodal Synchronization Detection | score | score | score | score | score | score | score | score | score |
| 13 | Long-Horizon Next-Action Forecasting | score | score | score | score | score | score | score | score | score |
| 14 | Long-Horizon Next-Subtask Forecasting | score | score | score | score | score | score | score | score | score |
| 15 | Interaction Text Prediction | score | score | score | score | proxy | proxy | score | score | score |
| 16 | Action-Object Relation Prediction | score | score | score | score | score | score | score | score | score |
| 17 | Future Object-Set Forecasting | score | score | score | score | score | score | score | score | score |
| 18 | IMU-to-Hand Pose Reconstruction | score | score | score | score | score | score | score | score | score |
| 19 | Camera-View Synchronization Retrieval | score | score | proxy | proxy | proxy | proxy | score | score | score |
| 20 | Time-to-Next-Transition Regression | score | score | score | score | score | score | score | score | score |

Sources and raw values are in `docs/data/task_method_20_result_matrix.json` and `docs/data/unified_task_model_radar.json`.