{"id":"W4408354411","doi":"10.1109/icassp49660.2025.10888241","title":"LAVCap: LLM-based Audio-Visual Captioning using Optimal Transport","year":2025,"lang":"en","type":"article","venue":"","topic":"Music and Audio Processing","field":"Computer Science","cited_by":6,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo","funders":"","keywords":"Closed captioning; Computer science; Audio visual; Speech recognition; Multimedia; Artificial intelligence; Image (mathematics)","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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.000217305,0.0001294744,0.0001520967,0.000144674,0.0002709398,0.0001434591,0.0003882962,0.00005913827,0.00007437975],"category_scores_gemma":[0.000009423641,0.0001203866,0.00007054228,0.0005544464,0.00004551012,0.0004129364,0.00005396284,0.000114492,0.00001237103],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0000486723,"about_ca_system_score_gemma":0.0003676136,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00006730715,"about_ca_topic_score_gemma":0.000007527169,"domain_scores_codex":[0.9989306,0.00002095219,0.0002213204,0.00035401,0.0001924742,0.0002806007],"domain_scores_gemma":[0.9995693,0.00003317318,0.00005353241,0.0002194136,0.00006783696,0.00005671837],"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":[0.00008872092,0.001203853,0.03203736,0.000744339,0.000237016,0.0003095412,0.003081204,0.3198308,0.08000033,0.1974979,0.007709897,0.357259],"study_design_scores_gemma":[0.0005412979,0.00003251088,0.002081719,0.0001506359,0.00001808545,0.000005948552,0.00004591004,0.9698848,0.02363024,0.0003881672,0.002951162,0.0002695402],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1291824,0.00006432278,0.8603166,0.0008895759,0.0002473871,0.00005786463,3.249412e-7,0.0002269188,0.009014559],"genre_scores_gemma":[0.7872152,6.747058e-7,0.2094508,0.002187887,0.0000429126,0.000003020315,0.000001609937,0.000005096303,0.001092742],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.6580328,"threshold_uncertainty_score":0.4909224,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01577899618749308,"score_gpt":0.2860186404324295,"score_spread":0.2702396442449365,"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."}}