{"id":"W4416551090","doi":"10.48550/arxiv.2504.12083","title":"Self-alignment of Large Video Language Models with Refined Regularized Preference Optimization","year":2025,"lang":"en","type":"preprint","venue":"ArXiv.org","topic":"Multimodal Machine Learning Applications","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Mitacs","keywords":"Spurious relationship; Regularization (linguistics); Preference; Set (abstract data type); Simple (philosophy); Optimization problem","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001992484,0.001292942,0.001074939,0.0005397365,0.0004194141,0.0008964199,0.001943999,0.00130526,0.002468449],"category_scores_gemma":[0.007887932,0.0006871799,0.001053802,0.0006371393,0.0007037793,0.001937379,0.001726339,0.002156937,0.00115606],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0009352334,"about_ca_system_score_gemma":0.00134829,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.005651402,"about_ca_topic_score_gemma":0.007261258,"domain_scores_codex":[0.998861,0.0004769995,0.00006305464,0.0003227718,0.0001623494,0.0001138311],"domain_scores_gemma":[0.9976598,0.001302424,0.0001964015,0.0002794656,0.0004207159,0.0001413027],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.0003382725,0.0001468586,0.001500294,0.0001437265,0.00008927831,0.0001524758,0.0002029729,0.8162304,0.008620819,0.009317501,0.00403192,0.1592255],"study_design_scores_gemma":[0.000007641155,0.00002420521,0.0000458664,0.00000281219,0.000002780682,0.000009516966,0.000008202886,0.996563,0.0007980455,0.002375715,0.0001576258,0.00000452468],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.02744972,0.0001479286,0.9695237,0.0001635382,0.00003751231,0.0000609796,0.0001336032,0.00172483,0.0007582357],"genre_scores_gemma":[0.6409659,0.0001712406,0.3503214,0.0006370789,0.00009159084,0.0004433997,0.001288892,0.0009555205,0.005124876],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.005651402,"threshold_uncertainty_score":0.01123703,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02632753695901735,"score_gpt":0.2761713555977348,"score_spread":0.2498438186387175,"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."}}