{"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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001693066,0.001796891,0.00116053,0.001185817,0.000889528,0.001135123,0.002342089,0.001582179,0.005786676],"category_scores_gemma":[0.006275106,0.0005967225,0.001222704,0.0008503011,0.001032062,0.002866322,0.003067402,0.003870615,0.004216507],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0006356603,"about_ca_system_score_gemma":0.001355879,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002032049,"about_ca_topic_score_gemma":0.004582121,"domain_scores_codex":[0.9986557,0.0003986014,0.00006851684,0.0004523364,0.0003184529,0.000106384],"domain_scores_gemma":[0.9978265,0.001069271,0.0001375054,0.0005382564,0.0003124728,0.0001159484],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.000656747,0.000387217,0.00136387,0.0002922775,0.00007173978,0.0002965392,0.0005362038,0.03749521,0.04121567,0.007416992,0.0279489,0.8823186],"study_design_scores_gemma":[0.00009058719,0.0002081637,0.0004698913,0.00003596217,0.00002383865,0.0001819035,0.0001439986,0.9375914,0.03143304,0.0165429,0.01321999,0.00005828791],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.01854792,0.0004276476,0.9541172,0.0003277225,0.0002583686,0.0001919495,0.0008904577,0.02376943,0.001469236],"genre_scores_gemma":[0.1824908,0.0002046545,0.8008963,0.0006963341,0.0002390893,0.0007807197,0.005886394,0.001821097,0.006984714],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.005786676,"threshold_uncertainty_score":0.0193584,"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."}}