{"id":"W4395443601","doi":"10.48550/arxiv.2404.15004","title":"TAXI: Evaluating Categorical Knowledge Editing for Language Models","year":2024,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Arizona State University","keywords":"Categorical variable; Computer science; Natural language processing; Artificial intelligence; Machine learning","routes":{"ca_aff":false,"ca_fund":false,"ca_venue":false,"about_ca":true,"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":[0.01322369,0.00237048,0.001281835,0.004201392,0.001309299,0.003394328,0.004883315,0.003333242,0.003611372],"category_scores_gemma":[0.08823721,0.0006161985,0.001653923,0.003480109,0.001619868,0.006733439,0.003156506,0.003673006,0.002377193],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002776038,"about_ca_system_score_gemma":0.002736964,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01695004,"about_ca_topic_score_gemma":0.02884763,"domain_scores_codex":[0.9841903,0.006199229,0.001506164,0.00360432,0.00397524,0.0005247254],"domain_scores_gemma":[0.9289085,0.04908547,0.002470647,0.01390368,0.004251149,0.001380612],"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":[0.003469211,0.002149091,0.05670268,0.005539439,0.001987854,0.0008611321,0.001747881,0.2106085,0.01637568,0.01287405,0.2740665,0.4136179],"study_design_scores_gemma":[0.00077495,0.001221192,0.01203416,0.0002866394,0.0002943601,0.001027771,0.0007845228,0.8911191,0.01930612,0.01959692,0.05334624,0.0002079799],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.5078102,0.01530144,0.1639439,0.00438926,0.002306227,0.001768169,0.1083571,0.1659793,0.03014437],"genre_scores_gemma":[0.5810255,0.001435949,0.2401886,0.001498388,0.0003854712,0.0007180767,0.1642348,0.005005965,0.005507334],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01695004,"threshold_uncertainty_score":0.06993443,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1197068682973257,"score_gpt":0.2748363199575214,"score_spread":0.1551294516601958,"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."}}