{"id":"W4402669690","doi":"10.18653/v1/2024.findings-acl.906","title":"TAXI: Evaluating Categorical Knowledge Editing for Language Models","year":2024,"lang":"en","type":"article","venue":"","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Arizona State University","keywords":"Computer science; Categorical variable; 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":[0.01246229,0.002258277,0.001190673,0.004135882,0.001231595,0.003060967,0.004471181,0.003101542,0.002964613],"category_scores_gemma":[0.07909601,0.0005371639,0.001548861,0.003264142,0.00148969,0.006058855,0.00280368,0.003273018,0.00183649],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002379846,"about_ca_system_score_gemma":0.002427877,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01665065,"about_ca_topic_score_gemma":0.02823349,"domain_scores_codex":[0.9840845,0.006393216,0.001654837,0.003449235,0.003915612,0.0005025652],"domain_scores_gemma":[0.9298581,0.04947979,0.002739187,0.0123389,0.004218181,0.001365831],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.003802917,0.00237985,0.07302657,0.005936247,0.002212805,0.00102191,0.001858481,0.2225192,0.01915785,0.01030064,0.2349292,0.4228542],"study_design_scores_gemma":[0.0007916858,0.001645211,0.01683526,0.0003207812,0.0003553445,0.001248336,0.0008766653,0.8863347,0.02184294,0.01557339,0.05394004,0.0002357253],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.5985717,0.01654706,0.137784,0.00386836,0.002221434,0.001743753,0.09297697,0.118486,0.02780081],"genre_scores_gemma":[0.6500206,0.001357661,0.2016697,0.001291616,0.0003511468,0.0005950296,0.1370791,0.003079241,0.004555874],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01665065,"threshold_uncertainty_score":0.06590766,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05127591607564936,"score_gpt":0.3814212288681417,"score_spread":0.3301453127924923,"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."}}