{"id":"W6966641450","doi":"10.48448/gfnv-vm46","title":"Universal-KD: Attention-based Output-Grounded Intermediate Layer Knowledge Distillation","year":2021,"lang":"en","type":"other","venue":"Underline Science Inc.","topic":"","field":"","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo","funders":"","keywords":"Distillation; Layer (electronics); Matching (statistics); Key (lock); Column (typography); Fractionating column","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":[0.0007866548,0.001433624,0.0009663289,0.0009337518,0.0006785169,0.001451177,0.003151319,0.00139593,0.008791932],"category_scores_gemma":[0.002740651,0.0005735658,0.001082833,0.001130075,0.00100651,0.004283353,0.004023575,0.002581447,0.00251451],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001124226,"about_ca_system_score_gemma":0.002034615,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.008434605,"about_ca_topic_score_gemma":0.01188429,"domain_scores_codex":[0.9993286,0.00008904924,0.00004628908,0.0002675719,0.0001349103,0.0001336042],"domain_scores_gemma":[0.9992445,0.0001881083,0.00006149077,0.0002987485,0.0001468846,0.00006036864],"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":[0.0002989398,0.0002027998,0.0009857597,0.0003371643,0.0001244992,0.0002399318,0.0003143277,0.1116734,0.01873781,0.02538371,0.01292727,0.8287744],"study_design_scores_gemma":[0.00003453076,0.00008358736,0.0003775334,0.00004036039,0.00005722358,0.0001039887,0.000074271,0.9327612,0.02418784,0.03356436,0.008682283,0.00003278189],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.01964892,0.00052809,0.961723,0.0003175273,0.000163986,0.00009562127,0.0004432848,0.01156205,0.005517401],"genre_scores_gemma":[0.5526193,0.0003609456,0.42809,0.0005835117,0.00007401926,0.000180251,0.002324921,0.000758631,0.01500822],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.008791932,"threshold_uncertainty_score":0.02941191,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03976833101735353,"score_gpt":0.3152315852717479,"score_spread":0.2754632542543943,"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."}}