{"id":"W4385570255","doi":"10.18653/v1/2023.acl-long.605","title":"f-Divergence Minimization for Sequence-Level Knowledge Distillation","year":2023,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":14,"is_retracted":false,"has_abstract":true,"ca_institutions":"Canadian Institute for Advanced Research; McGill University; University of Alberta","funders":"Alliance de recherche numérique du Canada; DeepMind; Natural Sciences and Engineering Research Council of Canada; Canadian Institute for Advanced Research","keywords":"Divergence (linguistics); Distillation; Sequence (biology); Computer science; Minification; Function (biology); Process (computing); Word (group theory); Decomposition; Artificial intelligence; Mathematics; Programming language","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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0001440981,0.00005732178,0.00005632689,0.00007360724,0.00009364299,0.00005062125,0.0003092515,0.00003147253,0.00001364743],"category_scores_gemma":[0.00008118181,0.00005320141,0.00002847007,0.0004101551,0.000008959883,0.000277546,0.0001112766,0.00002041722,0.000155545],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00002729069,"about_ca_system_score_gemma":0.00004520111,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000009750581,"about_ca_topic_score_gemma":0.00001282627,"domain_scores_codex":[0.9993666,0.0000131179,0.0001301757,0.0002441861,0.00009461535,0.000151341],"domain_scores_gemma":[0.9995188,0.00008274596,0.0000302512,0.0002351731,0.00009672947,0.00003627078],"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":[0.00000291073,0.00002009268,0.002456099,0.00004709208,0.000008749372,0.00000160743,0.001644833,0.01844503,0.001387854,0.7198611,0.01108948,0.2450351],"study_design_scores_gemma":[0.00008732227,0.00001028623,0.002162722,0.000005469324,0.000001282297,6.551326e-7,0.00001519228,0.9833369,0.0006038591,0.01036993,0.003329276,0.00007713491],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.005606094,0.00001003233,0.9900122,0.0005206326,0.0005281768,0.0001548561,0.000004565659,0.0003952005,0.002768205],"genre_scores_gemma":[0.7876219,0.000009469606,0.2058031,0.00005200942,0.00009317011,0.00003213484,0.0000172933,0.000005461092,0.006365474],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9648919,"threshold_uncertainty_score":0.2169491,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1834173923804379,"score_gpt":0.3351924403742882,"score_spread":0.1517750479938503,"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."}}