{"id":"W4412719203","doi":"10.1007/978-3-031-97992-7_63","title":"Evaluating LLMs for Effective Object Search in Video Data","year":2025,"lang":"en","type":"book-chapter","venue":"Lecture notes in networks and systems","topic":"Advanced Image and Video Retrieval Techniques","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":false,"ca_institutions":"Ontario Tech University","funders":"","keywords":"Object (grammar); Computer science; Geography; Artificial intelligence","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":[0.003906215,0.001022234,0.001932733,0.004731052,0.0005319734,0.0019795,0.001310598,0.002022367,0.004474421],"category_scores_gemma":[0.01615082,0.0002317074,0.0006810065,0.002542338,0.000541059,0.003413721,0.0009681848,0.0006329722,0.001148792],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001768004,"about_ca_system_score_gemma":0.001376177,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01009079,"about_ca_topic_score_gemma":0.0110053,"domain_scores_codex":[0.9971119,0.000918921,0.0003771672,0.0004020212,0.0009530844,0.00023689],"domain_scores_gemma":[0.9894474,0.008456329,0.0003398195,0.0005272403,0.0009486508,0.0002805302],"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":[0.004950869,0.0007369528,0.0132574,0.0007237424,0.0003957663,0.0001663957,0.0000904085,0.1151859,0.02332732,0.002021822,0.0105637,0.8285796],"study_design_scores_gemma":[0.00008391533,0.0008977397,0.002809142,0.00002941393,0.00008918736,0.0001387989,0.0001234437,0.9795216,0.01372219,0.001480866,0.001080412,0.00002334235],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.6246845,0.01816183,0.3341681,0.0008189497,0.0004084647,0.0004659209,0.00380242,0.01175239,0.005737452],"genre_scores_gemma":[0.8328891,0.001315332,0.1564763,0.0001954229,0.0002133971,0.0001392093,0.005172894,0.000262527,0.003335902],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01009079,"threshold_uncertainty_score":0.02065831,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.06372081166989486,"score_gpt":0.3714293954147437,"score_spread":0.3077085837448489,"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."}}