{"id":"W4410949880","doi":"10.1109/access.2025.3575440","title":"STAA: Spatio-Temporal Attention Attribution for Real-Time Interpreting Transformer-Based AI Video Models","year":2025,"lang":"en","type":"article","venue":"IEEE Access","topic":"Human Pose and Action Recognition","field":"Computer Science","cited_by":13,"is_retracted":false,"has_abstract":true,"ca_institutions":"Concordia University","funders":"Natural Sciences and Engineering Research Council of Canada","keywords":"Computer science; Transformer; Attribution; Artificial intelligence; Speech recognition; Psychology; Voltage; Electrical engineering; Social psychology; Engineering","routes":{"ca_aff":true,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0009349969,0.001293039,0.0005251172,0.001693708,0.0003133681,0.001280125,0.00164786,0.0006998986,0.006707697],"category_scores_gemma":[0.004751453,0.0004309872,0.001380221,0.0008587861,0.0004478436,0.001758162,0.001515295,0.001552366,0.002238814],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001091151,"about_ca_system_score_gemma":0.001129126,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01023774,"about_ca_topic_score_gemma":0.01791383,"domain_scores_codex":[0.9995314,0.00008929755,0.00002964052,0.0001750271,0.0001237047,0.00005094882],"domain_scores_gemma":[0.9989272,0.0004180196,0.0001216986,0.0002483027,0.0002117103,0.00007306982],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0004524444,0.0001974691,0.004609377,0.000269111,0.0002248961,0.0002662128,0.0004471954,0.07332774,0.02569589,0.01113261,0.01777741,0.8655996],"study_design_scores_gemma":[0.00001743253,0.00003921955,0.000829517,0.00001168854,0.00002461078,0.00006192196,0.00006226465,0.9818988,0.005670888,0.008693712,0.00267647,0.00001346632],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.01353025,0.0002615639,0.9638049,0.0001478054,0.00006937789,0.0001177291,0.0008919328,0.02002347,0.001152947],"genre_scores_gemma":[0.3967071,0.0003855792,0.5937597,0.0002007039,0.00008507192,0.0002216283,0.004134789,0.001019359,0.003486073],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.01023774,"threshold_uncertainty_score":0.02243948,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03212404749350114,"score_gpt":0.3268996864187976,"score_spread":0.2947756389252965,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}