{"id":"W6929111282","doi":"10.48448/1865-ga42","title":"MATE-KD: Masked Adversarial TExt, a Companion to 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 British Columbia","funders":"","keywords":"Adversarial system; Generator (circuit theory); Distillation; Benchmark (surveying); Key (lock); Set (abstract data type); Divergence (linguistics); Test set; Language model","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","insufficient_payload"],"consensus_categories":["insufficient_payload"],"category_scores_codex":[0.001107825,0.0005949911,0.0006594797,0.001811657,0.0003427304,0.0004456935,0.001249304,0.0003232633,0.006706375],"category_scores_gemma":[0.0004826155,0.0005811561,0.000112571,0.00384905,0.0009280507,0.0002632444,0.0007176149,0.000358743,0.01284294],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001076883,"about_ca_system_score_gemma":0.001509067,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0005446383,"about_ca_topic_score_gemma":0.004656423,"domain_scores_codex":[0.9954679,0.0001576946,0.0005367289,0.001541949,0.001382773,0.0009129455],"domain_scores_gemma":[0.997456,0.00008266146,0.0004092143,0.001123306,0.0004072612,0.0005215739],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"not_applicable","study_design_gemma":"not_applicable","study_design_scores_codex":[0.00008299323,0.0005992874,0.0004262153,0.0002170139,0.0001000995,0.0000452065,0.001060466,0.001047325,0.03805676,0.01480624,0.9312575,0.01230087],"study_design_scores_gemma":[0.0008931214,0.00009068509,0.001455778,0.0006137145,0.00007694522,0.0000202779,0.0002963202,0.01639688,0.0003581338,0.0002029254,0.9786208,0.0009743977],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"other","genre_gemma":"other","genre_scores_codex":[0.001121017,0.0005896006,0.007503098,0.0003560295,0.005481812,0.0016384,0.0006793521,0.001311169,0.9813195],"genre_scores_gemma":[0.3227297,0.0000239988,0.0302312,0.0002199045,0.003710297,0.00007280415,0.00155222,0.00191138,0.6395485],"genre_candidate":"other","genre_consensus":"other","teacher_disagreement_score":0.341771,"threshold_uncertainty_score":0.999664,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03073713003460477,"score_gpt":0.3330264999085704,"score_spread":0.3022893698739656,"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."}}