{"id":"W4402671356","doi":"10.18653/v1/2024.arabicnlp-1.18","title":"John vs. Ahmed: Debate-Induced Bias in Multilingual LLMs","year":2024,"lang":"en","type":"article","venue":"","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":5,"is_retracted":false,"has_abstract":true,"ca_institutions":"Artificial Intelligence in Medicine (Canada)","funders":"Alliance de recherche numérique du Canada; Social Sciences and Humanities Research Council of Canada; Natural Sciences and Engineering Research Council of Canada; Canada Research Chairs","keywords":"Political science; Computer science","routes":{"ca_aff":true,"ca_fund":true,"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":[0.02533729,0.0004848687,0.0004901661,0.001052141,0.002651944,0.004932683,0.0008793749,0.00198789,0.008426353],"category_scores_gemma":[0.1531487,0.0003935068,0.0003835282,0.0007837684,0.00386897,0.009852735,0.00483638,0.00294601,0.001298515],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002809007,"about_ca_system_score_gemma":0.001879565,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002266555,"about_ca_topic_score_gemma":0.002640674,"domain_scores_codex":[0.9630549,0.02825893,0.001419579,0.002170415,0.004185308,0.0009108671],"domain_scores_gemma":[0.842065,0.1280568,0.008949414,0.01036173,0.009374582,0.001192446],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.002533884,0.0003711369,0.07239778,0.00159828,0.000181319,0.002463758,0.2605487,0.003994363,0.01828039,0.4408905,0.02101604,0.1757239],"study_design_scores_gemma":[0.0003631206,0.0006654377,0.03265296,0.001885174,0.0003842435,0.003164232,0.1558562,0.05745972,0.04257199,0.4030901,0.3013961,0.0005106766],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.7641789,0.0009927454,0.08005169,0.01551949,0.0005397628,0.0002800442,0.0004276322,0.00073165,0.1372781],"genre_scores_gemma":[0.9884625,0.0001252423,0.007816914,0.0009219367,0.00004898277,0.00005666666,0.00008542868,0.0001945701,0.002287801],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.02533729,"threshold_uncertainty_score":0.133998,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03865088048033549,"score_gpt":0.3257001436754829,"score_spread":0.2870492631951474,"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."}}