{"id":"W6979255470","doi":"","title":"Relative Bias: A Comparative Framework for Quantifying Bias in LLMs","year":2025,"lang":"en","type":"article","venue":"ArXiv.org","topic":"Computational and Text Analysis Methods","field":"Social Sciences","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo","funders":"","keywords":"Ambiguity; Embedding; Transformation (genetics); Software deployment; Raising (metalworking); Sentence","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":[0.07647264,0.001775863,0.00140542,0.009701048,0.002303255,0.006503284,0.002795512,0.002927982,0.006309256],"category_scores_gemma":[0.3397835,0.0006238995,0.001657478,0.005850434,0.007017654,0.01298142,0.008198344,0.003428126,0.001032865],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.003184101,"about_ca_system_score_gemma":0.003175551,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002144922,"about_ca_topic_score_gemma":0.002677603,"domain_scores_codex":[0.9212452,0.05709216,0.003821325,0.005300236,0.01156698,0.0009740615],"domain_scores_gemma":[0.7094587,0.2302037,0.01929943,0.02278997,0.01657455,0.001673542],"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":[0.0006872625,0.0002101126,0.04534754,0.001263154,0.0009073847,0.0002771972,0.007509571,0.07705778,0.00512023,0.559534,0.005769806,0.296316],"study_design_scores_gemma":[0.00007269299,0.0004131324,0.00854482,0.0005273305,0.0001902417,0.0002562038,0.001768901,0.2666931,0.005056354,0.705535,0.0107614,0.000180875],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.03375255,0.0007923681,0.9549854,0.001243137,0.0001210285,0.0002965786,0.0005752976,0.000792297,0.007441335],"genre_scores_gemma":[0.584435,0.0003534028,0.410762,0.0005749274,0.0002128457,0.001174718,0.0007839619,0.0005446437,0.001158554],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.07647264,"threshold_uncertainty_score":0.4044308,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.5209298266243385,"score_gpt":0.5210722438492991,"score_spread":0.000142417224960556,"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."}}