{"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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.00386355,0.001269857,0.001826959,0.001365995,0.0007869498,0.001595149,0.002513851,0.002392586,0.002971218],"category_scores_gemma":[0.01109343,0.0004514217,0.001085934,0.001760136,0.001920775,0.00405794,0.002708633,0.003660746,0.001191859],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002057472,"about_ca_system_score_gemma":0.002599295,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004002642,"about_ca_topic_score_gemma":0.005051699,"domain_scores_codex":[0.9981388,0.000630498,0.0001314458,0.0004671051,0.0004605864,0.0001715564],"domain_scores_gemma":[0.9960854,0.002619234,0.0001952484,0.0005058813,0.00042825,0.0001659465],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.0001920559,0.0002217373,0.001137326,0.0002960116,0.00009251502,0.0001225931,0.0001926315,0.6269487,0.003577789,0.06713248,0.008430584,0.2916556],"study_design_scores_gemma":[0.00001265418,0.00003693191,0.0001164198,0.00001365478,0.000007069826,0.00003621254,0.0000187939,0.960269,0.001312792,0.03675284,0.001410439,0.00001313676],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.009950875,0.00048005,0.9872573,0.0003751272,0.00003806008,0.00004680269,0.0001514916,0.0005691671,0.001131185],"genre_scores_gemma":[0.420502,0.0008292997,0.5678107,0.0007892955,0.000212472,0.0003737802,0.001896821,0.0005403277,0.007045243],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.004002642,"threshold_uncertainty_score":0.02043259,"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."}}