{"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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001395059,0.0001010471,0.0002735112,0.0002002275,0.0004090415,0.00004678055,0.0001844602,0.0001040653,0.00006910264],"category_scores_gemma":[0.002272733,0.00009389752,0.0001398513,0.001333425,0.0001547492,0.0001784197,0.0000423446,0.000199823,0.00004183804],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00009620039,"about_ca_system_score_gemma":0.00017682,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0009625206,"about_ca_topic_score_gemma":0.003172769,"domain_scores_codex":[0.9985284,0.0005085632,0.0002825568,0.000265984,0.0001788861,0.0002355629],"domain_scores_gemma":[0.9946408,0.00493966,0.0001169454,0.0001066175,0.0001490722,0.0000469625],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.00002231601,0.00005276643,0.558461,0.000007328627,0.00006344114,0.000001181311,0.01479406,0.0002248062,0.00001559943,0.4236402,0.0001818207,0.002535524],"study_design_scores_gemma":[0.0004085053,0.00003501413,0.5744979,0.0002117349,0.00006105591,7.709604e-8,0.01392405,0.003094686,0.0001858482,0.3828318,0.02452131,0.0002280591],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.842729,0.000306779,0.1431831,0.002985606,0.0002955732,0.0002866708,0.000003196114,0.00004582927,0.01016425],"genre_scores_gemma":[0.9710441,0.00002111297,0.02662251,0.0003427658,0.0001024018,0.00004516628,0.000005594691,0.000004395912,0.001811935],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.1283151,"threshold_uncertainty_score":0.382903,"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."}}