{"id":"W7125605555","doi":"10.1109/cascon66301.2025.00103","title":"A Hybrid XAI Pipeline for Multimodal Video Understanding: From Transformers to LLMS","year":2025,"lang":"","type":"article","venue":"","topic":"Multimodal Machine Learning Applications","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Concordia University","funders":"","keywords":"Robustness (evolution); Scalability; Transformer; Software deployment; Pipeline (software); Exploit; Interoperability","routes":{"ca_aff":true,"ca_fund":false,"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":[0.001708817,0.00200721,0.001004867,0.002275376,0.0006006612,0.003293474,0.003012752,0.001086733,0.02246028],"category_scores_gemma":[0.00579423,0.0007452993,0.002333365,0.001167013,0.0006666614,0.004589919,0.003809556,0.002656804,0.01042683],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001538665,"about_ca_system_score_gemma":0.001323314,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.007684189,"about_ca_topic_score_gemma":0.006865972,"domain_scores_codex":[0.9993191,0.0001137337,0.00003992393,0.0002518544,0.0001895692,0.00008586108],"domain_scores_gemma":[0.998735,0.0004453452,0.00007608662,0.0003905604,0.0002653335,0.00008757616],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.0009255694,0.0002476401,0.003499883,0.0004406779,0.0001861401,0.0003692494,0.0007759744,0.02184819,0.02966102,0.02272264,0.0498917,0.8694312],"study_design_scores_gemma":[0.00006414911,0.0001839303,0.001134309,0.00007373533,0.0001008148,0.000252889,0.0003802894,0.8460171,0.06512628,0.03737304,0.04920545,0.00008802034],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.007135821,0.0002901984,0.8197157,0.0003662502,0.00008008583,0.0002100901,0.002218854,0.1669655,0.003017418],"genre_scores_gemma":[0.1730346,0.0005490218,0.7937429,0.00038713,0.0001086702,0.0004386489,0.0116058,0.01025156,0.009881775],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.02246028,"threshold_uncertainty_score":0.07513714,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03208313348499037,"score_gpt":0.3148403007469172,"score_spread":0.2827571672619268,"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."}}