{"id":"W4411630225","doi":"10.18653/v1/2024.eacl-long.41","title":"Examining Gender and Racial Bias in Large Vision–Language Models Using a Novel Dataset of Parallel Images","year":2024,"lang":"en","type":"article","venue":"","topic":"Categorization, perception, and language","field":"Psychology","cited_by":8,"is_retracted":false,"has_abstract":true,"ca_institutions":"National Research Council Canada","funders":"","keywords":"Computer science; Artificial intelligence; Gender bias; Racial bias; Natural language processing; Computer vision; Race (biology); Psychology; Sociology; Gender studies","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":["insufficient_payload"],"consensus_categories":[],"category_scores_codex":[0.0004151997,0.0001264712,0.0001746915,0.0002519031,0.00004121304,0.00005326772,0.00008095362,0.00009632143,0.003130269],"category_scores_gemma":[0.00001716388,0.0001105375,0.00002616815,0.0002447285,0.00004811056,0.0002942475,0.00006105594,0.0001069794,0.00001717115],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00001818462,"about_ca_system_score_gemma":0.00002856283,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00101488,"about_ca_topic_score_gemma":0.0003135219,"domain_scores_codex":[0.9989476,0.00005991804,0.0003121379,0.0003421307,0.000123035,0.0002151821],"domain_scores_gemma":[0.9995828,0.00007863669,0.00004177844,0.0002331222,0.00001934888,0.00004433078],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"qualitative","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0004003829,0.001730553,0.02559844,0.0008000296,0.0004052955,0.000472258,0.6291543,0.003573663,0.1631528,0.1373254,0.02154191,0.01584502],"study_design_scores_gemma":[0.01118561,0.000408942,0.2769981,0.0004087888,0.0004095734,0.0005761003,0.2068643,0.486419,0.0006421709,0.003731088,0.01001109,0.00234523],"study_design_candidate":"qualitative","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.7818139,0.002134799,0.2090442,0.0000128806,0.0002374782,0.0001796212,0.001485231,0.00006316671,0.005028783],"genre_scores_gemma":[0.9947913,0.00003015808,0.003267286,0.0001091651,0.0001129179,0.000006423,0.001043915,0.00002363666,0.0006152478],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.4828453,"threshold_uncertainty_score":0.997781,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1388596897583776,"score_gpt":0.40015976839756,"score_spread":0.2613000786391824,"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."}}