{"id":"W4385570159","doi":"10.18653/v1/2023.findings-acl.714","title":"A Study on Knowledge Distillation from Weak Teacher for Scaling Up Pre-trained Language Models","year":2023,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":3,"is_retracted":false,"has_abstract":true,"ca_institutions":"Kootenay Association for Science & Technology","funders":"","keywords":"Initialization; Computer science; Artificial intelligence; Weighting; Distillation; Machine learning; Language model; Domain (mathematical analysis); Natural language processing; Mathematics","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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0003260107,0.0001113872,0.0001291074,0.00009040038,0.0000884645,0.0001004755,0.0003871196,0.00004434748,0.000009906956],"category_scores_gemma":[0.00003453918,0.00009386885,0.00005166558,0.0002220726,0.000005906385,0.0002189517,0.000156889,0.0000688319,0.00005629209],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00003726122,"about_ca_system_score_gemma":0.00002527504,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001231588,"about_ca_topic_score_gemma":0.00005615098,"domain_scores_codex":[0.9989491,0.00004858014,0.0001914104,0.0004330356,0.0001695766,0.0002082971],"domain_scores_gemma":[0.9992265,0.000171619,0.00003588849,0.0004786602,0.00003568429,0.00005171139],"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":[0.00009108364,0.0009740407,0.004204936,0.00005324556,0.0001943457,0.00001745162,0.279709,0.1379954,0.002761163,0.1314619,0.004442573,0.4380948],"study_design_scores_gemma":[0.0005265691,0.00004646005,0.004890626,0.00001381952,0.000006121124,1.541115e-7,0.001558801,0.9867971,0.0001441285,0.005837478,0.00005736698,0.0001214025],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.4225604,0.00001194527,0.5735072,0.0001148954,0.0002167107,0.0003420413,0.000003011992,0.0004002973,0.002843441],"genre_scores_gemma":[0.9809014,3.337002e-7,0.01258979,0.00002634401,0.0001747864,0.00006732444,0.00001036912,0.00001290577,0.006216768],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.8488017,"threshold_uncertainty_score":0.3827861,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.0692850980979679,"score_gpt":0.3390118085480475,"score_spread":0.2697267104500796,"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."}}