{"id":"W2947280697","doi":"10.48550/arxiv.1905.01347","title":"Auditing ImageNet: Towards a Model-driven Framework for Annotating Demographic Attributes of Large-Scale Image Datasets","year":2019,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Face recognition and analysis","field":"Computer Science","cited_by":15,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo","funders":"","keywords":"Computer science; Annotation; Scale (ratio); Audit; Demographics; Artificial intelligence; Machine learning; Code (set theory); Point (geometry); Variety (cybernetics); Data science; Set (abstract data type); Cartography","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.008984737,0.001618292,0.001158578,0.004204405,0.001075373,0.002223751,0.003134748,0.001954882,0.001598297],"category_scores_gemma":[0.02632138,0.0008485233,0.001059668,0.002868605,0.001120965,0.004540192,0.003560084,0.003450317,0.002549958],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002072517,"about_ca_system_score_gemma":0.00224408,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01464381,"about_ca_topic_score_gemma":0.03845957,"domain_scores_codex":[0.9952992,0.001866368,0.0002327916,0.00137058,0.000940509,0.0002905162],"domain_scores_gemma":[0.9906294,0.003017666,0.001156051,0.002567809,0.002150033,0.0004791514],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0007935389,0.0009765628,0.1099205,0.0006366211,0.0004604153,0.0005310744,0.001265175,0.08530156,0.01553899,0.02866045,0.1995614,0.5563538],"study_design_scores_gemma":[0.00005277582,0.0001637188,0.02206553,0.0001470741,0.00006127821,0.0004471282,0.0005003412,0.8720304,0.01073996,0.04834865,0.04534502,0.00009812263],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.07238706,0.0009093261,0.8677754,0.002253206,0.0003909786,0.0008925511,0.02946493,0.02108998,0.004836562],"genre_scores_gemma":[0.3142173,0.000650044,0.5826633,0.00179954,0.0003858904,0.002027391,0.09164339,0.001682703,0.004930321],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.01464381,"threshold_uncertainty_score":0.04751635,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.0642147363907538,"score_gpt":0.2339749395754144,"score_spread":0.1697602031846606,"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."}}