{"id":"W4404672160","doi":"10.1007/978-3-031-73024-5_10","title":"Markov Knowledge Distillation: Make Nasty Teachers Trained by Self-undermining Knowledge Distillation Fully Distillable","year":2024,"lang":"en","type":"book-chapter","venue":"Lecture notes in computer science","topic":"Machine Learning and Data Classification","field":"Computer Science","cited_by":2,"is_retracted":false,"has_abstract":false,"ca_institutions":"University of Waterloo","funders":"","keywords":"Distillation; Computer science; Markov chain; Markov process; Artificial intelligence; Machine learning; Chromatography; Chemistry; Mathematics; Statistics","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":["metaepi_narrow","scholarly_communication"],"consensus_categories":[],"category_scores_codex":[0.001530298,0.0008797704,0.0006916113,0.0008520324,0.0005854636,0.001491642,0.002593001,0.0005077593,0.00003669098],"category_scores_gemma":[0.0002217446,0.0008151941,0.0001958804,0.001309958,0.0005338681,0.0006616663,0.001283905,0.001367628,0.0003071067],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.000873238,"about_ca_system_score_gemma":0.0008048204,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00002015472,"about_ca_topic_score_gemma":0.0001604063,"domain_scores_codex":[0.9946139,0.000120569,0.0008963069,0.002610943,0.0009021444,0.0008561079],"domain_scores_gemma":[0.9964117,0.000791474,0.0004453479,0.001724934,0.000301429,0.0003250512],"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":[0.000008511692,0.00008445235,0.0002532623,0.0002197147,0.00003930339,0.00002839736,0.004126715,0.002658974,0.00008143112,0.03899654,0.0008239798,0.9526787],"study_design_scores_gemma":[0.0003142878,0.0001433708,0.0009525529,0.0005124673,0.00003824452,0.00006730403,0.000001530829,0.8405368,0.00004059774,0.02881814,0.1275425,0.001032289],"study_design_candidate":"design_other","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.0001208442,0.002535491,0.94311,0.001248372,0.002466766,0.0004811377,0.00005370767,0.000871302,0.04911237],"genre_scores_gemma":[0.7952172,0.0001359219,0.1831152,0.0002477845,0.001618816,0.00004575299,0.00043585,0.0001695044,0.01901404],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9516464,"threshold_uncertainty_score":0.9995449,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01415122816492169,"score_gpt":0.2643206319185038,"score_spread":0.2501694037535821,"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."}}