{ "train": { "n_samples": 512, "label_counts_by_class_index": [ 256, 0, 0, 0, 0, 0, 0, 73, 75, 61, 47, 0 ], "label_counts_named": { "stable": 256, "pre_slip": 0, "pre_drop": 0, "collision_precursor": 0, "fall_precursor": 0, "navigation_dead_end": 0, "wrong_click": 0, "destructive_action": 73, "workflow_divergence": 75, "form_submission_failure": 61, "stale_context_action": 47, "invalid_tool_selection": 0 }, "active_class_indices": [ 0, 7, 8, 9, 10 ], "active_classes_named": [ "stable", "destructive_action", "workflow_divergence", "form_submission_failure", "stale_context_action" ], "active_class_count": 5, "schema_class_count": 12, "dead_class_count": 7, "stable_fraction": 0.5, "pair_count": 256, "state_identical_pairs": 256, "context_identical_pairs": 256, "action_identical_pairs": 256, "last_frame_identical_pairs": 256, "frame_bag_identical_pairs": 256, "per_step_mean_frame_l1": [ 0.0, 0.020954610779881477, 0.02762388251721859, 0.025509674102067947, 0.025776954367756844, 0.02479478530585766, 0.023527709767222404, 0.0 ], "per_step_max_frame_l1": [ 0.0, 0.11713975667953491, 0.12950395047664642, 0.12264633178710938, 0.10307323932647705, 0.12950395047664642, 0.11706987768411636, 0.0 ], "per_step_signal_active_step_count": 6, "per_step_signal_active_steps": [ 0, 1, 1, 1, 1, 1, 1, 0 ], "early_frame_signal_to_late_ratio": 0.9998477697372437, "per_family_step_signal": { "delayed_consequence": [ 0.0, 0.02646506577730179, 0.02175237610936165, 0.02136673405766487, 0.019724978134036064, 0.019385861232876778, 0.01901235803961754, 0.0 ], "persistence_then_reversal": [ 0.0, 0.014361280016601086, 0.02765071392059326, 0.03327198699116707, 0.03650447353720665, 0.022305164486169815, 0.028729110956192017, 0.0 ], "order_reversal": [ 0.0, 0.022411763668060303, 0.029535019770264626, 0.02044849842786789, 0.024918952956795692, 0.024966586381196976, 0.02770298533141613, 0.0 ], "regression_after_progress": [ 0.0, 0.019186286255717278, 0.030522391200065613, 0.028941072523593903, 0.02481175772845745, 0.030586998909711838, 0.019876690581440926, 0.0 ] }, "per_family_count": { "delayed_consequence": 61, "persistence_then_reversal": 47, "order_reversal": 73, "regression_after_progress": 75 }, "temporal_order_v2_event_step_pair_counts": { "1,2": 10, "1,3": 15, "4,5": 16, "3,5": 19, "2,5": 21, "2,6": 18, "2,3": 23, "3,4": 15, "4,6": 19, "2,4": 22, "1,5": 17, "1,4": 18, "1,6": 11, "5,6": 14, "3,6": 18 }, "has_pair_metadata": true }, "test": { "n_samples": 256, "label_counts_by_class_index": [ 128, 0, 0, 0, 0, 0, 0, 38, 36, 24, 30, 0 ], "label_counts_named": { "stable": 128, "pre_slip": 0, "pre_drop": 0, "collision_precursor": 0, "fall_precursor": 0, "navigation_dead_end": 0, "wrong_click": 0, "destructive_action": 38, "workflow_divergence": 36, "form_submission_failure": 24, "stale_context_action": 30, "invalid_tool_selection": 0 }, "active_class_indices": [ 0, 7, 8, 9, 10 ], "active_classes_named": [ "stable", "destructive_action", "workflow_divergence", "form_submission_failure", "stale_context_action" ], "active_class_count": 5, "schema_class_count": 12, "dead_class_count": 7, "stable_fraction": 0.5, "pair_count": 128, "state_identical_pairs": 128, "context_identical_pairs": 128, "action_identical_pairs": 128, "last_frame_identical_pairs": 128, "frame_bag_identical_pairs": 128, "per_step_mean_frame_l1": [ 0.0, 0.028466641902923584, 0.02343955636024475, 0.019437870010733604, 0.025724371895194054, 0.028658127412199974, 0.025457071140408516, 0.0 ], "per_step_max_frame_l1": [ 0.0, 0.12556621432304382, 0.12058135122060776, 0.10886863619089127, 0.12556621432304382, 0.1183348074555397, 0.11915408819913864, 0.0 ], "per_step_signal_active_step_count": 6, "per_step_signal_active_steps": [ 0, 1, 1, 1, 1, 1, 1, 0 ], "early_frame_signal_to_late_ratio": 0.8935928344726562, "per_family_step_signal": { "delayed_consequence": [ 0.0, 0.029185175895690918, 0.014790527522563934, 0.02266974188387394, 0.03251323103904724, 0.018067987635731697, 0.021109655499458313, 0.0 ], "order_reversal": [ 0.0, 0.029604695737361908, 0.030482446774840355, 0.0185006782412529, 0.022353418171405792, 0.0350334458053112, 0.02550520747900009, 0.0 ], "persistence_then_reversal": [ 0.0, 0.03327863663434982, 0.020518233999609947, 0.022953815758228302, 0.029448146000504494, 0.028634393587708473, 0.015502340160310268, 0.0 ], "regression_after_progress": [ 0.0, 0.022776348516345024, 0.024205856025218964, 0.015342595055699348, 0.021653534844517708, 0.029008494690060616, 0.036600153893232346, 0.0 ] }, "per_family_count": { "delayed_consequence": 24, "order_reversal": 38, "persistence_then_reversal": 30, "regression_after_progress": 36 }, "temporal_order_v2_event_step_pair_counts": { "1,4": 9, "1,2": 7, "2,5": 9, "2,4": 9, "1,3": 10, "1,5": 12, "5,6": 12, "1,6": 9, "3,5": 7, "4,6": 9, "4,5": 10, "3,4": 7, "2,3": 5, "2,6": 9, "3,6": 4 }, "has_pair_metadata": true }, "train_path": "runs/benchmark_v2/data/software_synthetic_n512_seed910.npz", "test_path": "runs/benchmark_v2/data/software_synthetic_n256_seed2910.npz" }