{"id":"W6976597069","doi":"10.60692/ck6mt-jht14","title":"f-Divergence Minimization for Sequence-Level Knowledge Distillation","year":2023,"lang":"en","type":"article","venue":"Greater South Information System","topic":"Topic Modeling","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Canadian Institute for Advanced Research; McGill University; University of Alberta","funders":"","keywords":"Distillation; Divergence (linguistics); Minification; Process (computing); Decomposition; Natural language","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.003827699,0.001219981,0.001762332,0.001368945,0.0007930717,0.001576196,0.002421502,0.002311929,0.002910115],"category_scores_gemma":[0.0109946,0.0004379344,0.001052511,0.001732404,0.001858659,0.003895056,0.002622223,0.00355403,0.001103679],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002137576,"about_ca_system_score_gemma":0.002551809,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004251463,"about_ca_topic_score_gemma":0.00513508,"domain_scores_codex":[0.9982168,0.0006142581,0.0001251383,0.0004360883,0.0004452262,0.0001623854],"domain_scores_gemma":[0.9962081,0.002581049,0.0001836188,0.0004644205,0.0004079065,0.0001548848],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0001721361,0.000204396,0.00114301,0.0002619496,0.00008755472,0.000115845,0.0001781744,0.6496397,0.003047674,0.0684181,0.008015264,0.2687162],"study_design_scores_gemma":[0.00001076785,0.00003330331,0.0001148876,0.00001237305,0.000006189413,0.00003050109,0.00001642189,0.9622152,0.001064552,0.03520898,0.001274876,0.00001183384],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.01000002,0.0004396762,0.98729,0.0003841457,0.00003643658,0.00004648764,0.0001439562,0.0004836041,0.001175678],"genre_scores_gemma":[0.4101383,0.0007616034,0.578282,0.000734774,0.0002058028,0.0003670804,0.001749235,0.000481974,0.007279162],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.004251463,"threshold_uncertainty_score":0.02024311,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1529844675221668,"score_gpt":0.2700700095088802,"score_spread":0.1170855419867133,"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."}}