加入实验输出（批次 10/14）

deep_learning/Q1/outputs/method_comparison/README.md

@@ -0,0 +1,7 @@
+# Q1 method comparison
+
+This folder contains grouped cross-validation results for M1–M4. M1/M2 are fixed rules; M3/M4 are trained without emotion labels. All learned models and probes use training-fold-only feature normalization, and folds are grouped by `video_id`.
+
+`alignment_metrics.csv` reports expected-time trajectories, monotonicity, attention entropy, and width diagnostics. `retrieval_probe_metrics.csv` uses a separately trained linear projection probe; its grid-index positives are not independent temporal ground truth. `reconstruction_probe_metrics.csv` reports held-out masked reconstruction error in training-fold standardized feature units. `frozen_emotion_probe_metrics.csv` is a small-sample downstream utility check.
+
+No human event-time annotation is present, so the results cannot establish direct human alignment accuracy. See `run_manifest.json` for parameters, seeds, device, and interpretation limits.

deep_learning/Q1/outputs/method_comparison/alignment_summary.csv

@@ -0,0 +1,13 @@
+method,modality,n,mvr_mean,mvr_std,normalized_entropy_mean,normalized_entropy_std,width80_source_positions_mean,width80_source_positions_std,expected_time_start_s_mean,expected_time_start_s_std,expected_time_end_s_mean,expected_time_end_s_std,trajectory_span_fraction_mean,trajectory_span_fraction_std
+M1,text,100,0.0,0.0,0.0008249471336603165,0.005233484169432736,1.004799998998642,0.0312849358458409,0.373650000790672,0.28760723340035366,7.3229000100512325,4.567867965796829,0.8612861529795919,0.07875035859337948
+M1,audio,100,0.0,0.0,0.35824004098773005,0.05432258900057586,12.056200053691864,7.307551875172266,0.3177500062027157,0.2794568898376313,7.441361117547106,4.533387053269832,0.8937758523714728,0.06338502509117576
+M1,vision,100,0.0,0.0,0.02091189981205389,0.03423249778642653,1.144999988079071,0.27983220085257804,0.30618889875375566,0.2956530091396046,7.43663269007148,4.535796288658362,0.8956127409730107,0.0656268624353691
+M2,text,100,0.0,0.0,0.003971234038472176,0.010095580232451877,1.0215999960899353,0.060615033764357146,0.3753062509072731,0.2858492171110105,7.3229000100512325,4.567867965796829,0.8611038638930768,0.07866203063075565
+M2,audio,100,0.0,0.0,0.39199826538562776,0.04856653297051989,12.777399990558624,7.437441491467199,0.08135000219190694,0.046356409936463026,7.716148159798975,4.594836781525238,0.968455815287307,0.009448427161791137
+M2,vision,100,0.0,0.0,0.015714890849776567,0.04120196183036711,1.108399999141693,0.3072087474461631,0.024000000567226686,0.0514830577428781,7.7077327980137875,4.617512750249008,0.9744537002407014,0.016307502074103464
+M3,text,300,0.0,0.0,0.0004981429129838944,0.004964740878146852,1.0029999995231629,0.02989949245257891,0.37365000079067195,0.28664372237282515,7.320650003693404,4.545633324884766,0.8611844960658346,0.07838045740132776
+M3,audio,300,0.01857142824679613,0.028241165293696132,0.992265993754069,0.00445452263162519,607.8159977213542,365.5015286317014,3.883805515007902,2.328319755482711,3.8684255829138965,2.2993181536785716,-0.001562372346719104,0.016018112979249148
+M3,vision,300,0.06843537279715141,0.06722724739800441,0.9827406370639801,0.021948411615917433,31.175333293279014,18.019629268682312,3.8615792202155004,2.266734235588232,3.834497800808173,2.259828297195092,-0.004611149926980335,0.0433981312652511
+M4,text,300,0.0,0.0,0.9999777108430863,4.937972466967483e-06,15.891133365631104,8.431963588888667,3.83314954836363,2.345445678904365,3.8348252384682473,2.3463801144674252,0.00020046850045522113,0.0014442569299196843
+M4,audio,300,0.0,0.0,0.9999900078773498,3.4153207662384735e-06,618.1359326680501,369.4452964156476,3.867822667326802,2.309874647070878,3.868319556290074,2.3104860442974764,4.607309897740703e-05,0.0006428369960135046
+M4,vision,300,0.0,0.0,0.9999960112571716,4.416330070963206e-06,32.02326666514079,18.480166711480305,3.8529649544991638,2.3080516925693866,3.8542076376005214,2.3106234287813843,0.00010548800230026034,0.001267114074095688