{"id":"W4416183485","doi":"10.1109/mipr67560.2025.00054","title":"Mitigating Image Captioning Hallucinations in Vision-Language Models","year":2025,"lang":"","type":"article","venue":"","topic":"Multimodal Machine Learning Applications","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo","funders":"","keywords":"Closed captioning; Retraining; Normalization (sociology); Inference; Reliability (semiconductor); Adaptation (eye); Test data; Reduction (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":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0009445267,0.0003289883,0.0003352663,0.0008246992,0.0006539383,0.0008321695,0.001295391,0.0001692453,0.000250803],"category_scores_gemma":[0.0004114844,0.0003737735,0.0001278828,0.002669375,0.000145374,0.001457678,0.0008472184,0.0008444826,0.0002060027],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002864664,"about_ca_system_score_gemma":0.0003523288,"about_ca_topic_candidate":true,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.007740235,"about_ca_topic_score_gemma":0.0006029321,"domain_scores_codex":[0.9967522,0.0003003321,0.0008790611,0.001075027,0.0003857074,0.0006076769],"domain_scores_gemma":[0.9976547,0.0005675015,0.0002078124,0.001190383,0.000243593,0.0001360031],"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":[0.000003426027,0.0003847645,0.001630614,0.00009655406,0.00002843215,0.0000162224,0.009503652,0.2234188,0.007740069,0.6073005,0.0003070689,0.1495699],"study_design_scores_gemma":[0.0005846454,0.00002154385,0.02346161,0.0003649395,0.00001299206,0.000003849733,0.0006392946,0.954853,0.0004245469,0.01928696,0.00004237776,0.000304228],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.06309055,0.0003692963,0.8157803,0.01113463,0.0002110875,0.0005582777,0.000005228252,0.0002576633,0.108593],"genre_scores_gemma":[0.7613578,0.00001496062,0.2347365,0.0005232276,0.00003355961,0.0001086165,0.000009138944,0.0000161023,0.003199998],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.7314342,"threshold_uncertainty_score":0.9998714,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.009100288445900715,"score_gpt":0.3229183336735161,"score_spread":0.3138180452276154,"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."}}