{"id":"W4415214274","doi":"10.1016/j.neucom.2025.131217","title":"Sycophancy in vision-language models: A systematic analysis and an inference-time mitigation framework","year":2025,"lang":"en","type":"article","venue":"Neurocomputing","topic":"Multimodal Machine Learning Applications","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":false,"ca_institutions":"","funders":"National Key Research and Development Program of China; Youth Innovation Promotion Association of the Chinese Academy of Sciences; Chinese Academy of Sciences; National Natural Science Foundation of China; Canadian Anesthesiologists' Society","keywords":"Sensitivity (control systems); Inference; Filter (signal processing); Trustworthiness; Decoding methods; Polarity (international relations); Path (computing); Mechanism (biology)","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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0004289409,0.0001467965,0.0003310758,0.0006335301,0.0001593226,0.0002915589,0.0005624528,0.00007335473,0.000002273488],"category_scores_gemma":[0.0003194593,0.0001463165,0.00005159244,0.002477616,0.00002136594,0.0003709903,0.0002896145,0.0002943824,0.00001059313],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00003152013,"about_ca_system_score_gemma":0.00003371143,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0002297336,"about_ca_topic_score_gemma":0.00001586684,"domain_scores_codex":[0.9983226,0.00031435,0.0004093796,0.000558364,0.0001823454,0.0002129535],"domain_scores_gemma":[0.9981313,0.0009575966,0.0001527926,0.0006398484,0.00005197084,0.00006648578],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.000007274348,0.0003856029,0.04713511,0.003547005,0.0001994916,0.00003204727,0.01525918,0.6878718,0.003889171,0.1640854,0.000008513112,0.07757939],"study_design_scores_gemma":[0.0001062339,0.00001884318,0.04448877,0.0008392567,0.00003845906,0.000001334503,0.0000317886,0.945417,0.00003668286,0.008915423,2.834235e-7,0.0001058851],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.4119566,0.00007157703,0.5870768,0.0002078935,0.00002774122,0.0002546656,3.346686e-7,0.0001541815,0.0002501759],"genre_scores_gemma":[0.9042035,0.000001822331,0.0954662,0.0002502771,0.00001746971,0.00003719626,0.000003050046,0.000006416047,0.00001403106],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.4922469,"threshold_uncertainty_score":0.5966613,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.007109514855624758,"score_gpt":0.3227593656927562,"score_spread":0.3156498508371314,"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."}}