{"id":"W4416817077","doi":"10.1016/j.compbiomed.2025.111347","title":"Med-VCD: Mitigating hallucination for medical large vision language models through visual contrastive decoding","year":2025,"lang":"en","type":"article","venue":"Computers in Biology and Medicine","topic":"Multimodal Machine Learning Applications","field":"Computer Science","cited_by":4,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Calgary","funders":"","keywords":"Decoding methods; Context (archaeology); Redundancy (engineering); Hallucinating; Modalities; Language model","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":[0.0008740314,0.0001210268,0.0002585195,0.0001486404,0.0001683922,0.00001248024,0.0003273766,0.0001502109,0.000004578971],"category_scores_gemma":[0.0005045275,0.0000984892,0.00002127943,0.0002452959,0.0001695918,0.0001372278,0.0002145567,0.0002325286,7.803382e-7],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00003828411,"about_ca_system_score_gemma":0.00004780056,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001290539,"about_ca_topic_score_gemma":0.0000405972,"domain_scores_codex":[0.998832,0.0001269324,0.000294082,0.0004073479,0.00009702837,0.0002426516],"domain_scores_gemma":[0.9983104,0.001336923,0.00008073413,0.000154867,0.00005650018,0.00006053424],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00001951784,0.00008018305,0.003107551,0.00005076713,0.00002636098,0.000005257469,0.003796558,0.0002030366,0.000954794,0.6961957,0.0003307452,0.2952295],"study_design_scores_gemma":[0.001940063,0.0001516443,0.005029802,0.000297678,0.000006886228,0.000006071185,0.0001210777,0.9528896,0.00007391527,0.03900895,0.0003872733,0.00008700273],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.04482239,0.0007007646,0.9416897,0.01137117,0.0003731504,0.0002982604,0.000001941645,0.00008530341,0.0006572955],"genre_scores_gemma":[0.892086,0.00005153714,0.1050746,0.002594559,0.00009692553,0.0000465879,0.00003332476,0.000004057294,0.00001232472],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9526866,"threshold_uncertainty_score":0.4016273,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01191472068424883,"score_gpt":0.4048271390593892,"score_spread":0.3929124183751403,"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."}}