{"id":"W4389518772","doi":"10.18653/v1/2023.emnlp-main.323","title":"PromptMix: A Class Boundary Augmentation Method for Large Language Model Distillation","year":2023,"lang":"en","type":"article","venue":"","topic":"Speech Recognition and Synthesis","field":"Computer Science","cited_by":11,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo; McGill University; Canadian Institute for Advanced Research","funders":"","keywords":"Class (philosophy); Distillation; Computer science; Boundary (topology); Programming language; Artificial intelligence; Mathematics; Chemistry; Chromatography; Mathematical analysis","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":[0.0004868068,0.00007271936,0.00008583916,0.0001200291,0.0001348545,0.000140173,0.0001707653,0.00003919162,0.00005275451],"category_scores_gemma":[0.00007427225,0.00006427646,0.00006247377,0.0003074792,0.000005187471,0.0003062045,0.00005502299,0.0000310802,0.0001558981],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00003099798,"about_ca_system_score_gemma":0.00004185314,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000008091257,"about_ca_topic_score_gemma":0.00003123636,"domain_scores_codex":[0.999226,0.00004148892,0.0001374998,0.0002398521,0.0001660503,0.0001890979],"domain_scores_gemma":[0.9995122,0.0001607202,0.00004223796,0.0001851837,0.00005361264,0.00004607964],"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":[0.00002067271,0.00008267468,0.00005907077,0.00005529177,0.00003310205,0.000006041322,0.002995356,0.0004505207,0.0102271,0.1599961,0.02113991,0.8049342],"study_design_scores_gemma":[0.0003057151,0.00001607537,0.0002870868,0.000004655854,0.000004977841,0.000001895568,0.000184024,0.9776,0.007079818,0.01182309,0.002599151,0.0000935412],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.004113917,0.000006367184,0.9882335,0.001271501,0.0001212551,0.0003088293,0.00002707999,0.0005402949,0.005377233],"genre_scores_gemma":[0.1041436,0.000003555397,0.8888209,0.0007455946,0.00005056614,0.0001557165,0.00009107388,0.00001201096,0.005976983],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.9771494,"threshold_uncertainty_score":0.2621118,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03956160061057078,"score_gpt":0.3467287036544414,"score_spread":0.3071671030438706,"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."}}