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