{"id":"W4399453597","doi":"10.48550/arxiv.2406.04220","title":"BEADs: Bias Evaluation Across Domains","year":2024,"lang":"en","type":"preprint","venue":"ArXiv.org","topic":"Topic Modeling","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Government of Canada; Canadian Institute for Advanced Research","keywords":"Environmental science","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.02552653,0.003464291,0.002039189,0.003834754,0.001906915,0.003950456,0.003277662,0.003609376,0.006041033],"category_scores_gemma":[0.06353515,0.0007728427,0.002223669,0.002776356,0.001950973,0.004923367,0.005542905,0.003912132,0.006160946],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002489487,"about_ca_system_score_gemma":0.003112912,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.008421136,"about_ca_topic_score_gemma":0.01006288,"domain_scores_codex":[0.9778901,0.01331157,0.001367477,0.003828055,0.002877583,0.0007251726],"domain_scores_gemma":[0.9624047,0.02313087,0.001260385,0.008102426,0.004111471,0.0009900954],"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":[0.003840521,0.001049409,0.05721708,0.002434737,0.001584721,0.0005175964,0.001134269,0.1726852,0.007787799,0.01785543,0.26334,0.4705533],"study_design_scores_gemma":[0.0008808947,0.000713259,0.01047969,0.0006007827,0.0004884101,0.0006485709,0.000665226,0.8425882,0.02217447,0.05707743,0.06347428,0.0002087521],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.2479601,0.01792718,0.5927596,0.006477769,0.002685995,0.001488538,0.03968617,0.06756082,0.02345384],"genre_scores_gemma":[0.6501753,0.002068239,0.2311437,0.002990492,0.0008340175,0.00132366,0.0964609,0.006222159,0.008781351],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.02552653,"threshold_uncertainty_score":0.1349988,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1884305838164817,"score_gpt":0.3731990703844707,"score_spread":0.184768486567989,"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."}}