{"id":"W4378505314","doi":"10.48550/arxiv.2305.15096","title":"Dynamic Masking Rate Schedules for MLM Pretraining","year":2023,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Speech Recognition and Synthesis","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Samsung; York University; Samsung Advanced Institute of Technology","keywords":"Masking (illustration); Computer science; Speedup; Transformer; Pareto principle; Schedule; Speech recognition; Real-time computing; Parallel computing; Mathematical optimization; Mathematics; Operating system; Engineering; Electrical engineering; Voltage","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0005498321,0.0002813614,0.0003160271,0.000408737,0.0002182585,0.0002249394,0.001479638,0.0002784379,0.00003109018],"category_scores_gemma":[0.0001879307,0.0003403304,0.000306977,0.0004841679,0.00006598953,0.0002925521,0.001081638,0.0004046226,0.000201546],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001622937,"about_ca_system_score_gemma":0.0001800049,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00002243875,"about_ca_topic_score_gemma":0.00004913419,"domain_scores_codex":[0.9980848,0.0001218483,0.000200131,0.001110293,0.0000746187,0.0004083289],"domain_scores_gemma":[0.9981986,0.0004662161,0.0002220816,0.0008250186,0.0001535429,0.0001345239],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0001497605,0.0002603671,0.002022867,0.001028855,0.001026569,0.001325255,0.001429972,0.1524745,0.0005765971,0.7486663,0.0008631643,0.09017584],"study_design_scores_gemma":[0.0002949543,0.00002203784,0.0006239251,0.0001792782,0.00004606177,0.000001973453,0.000158237,0.8766989,0.0002617593,0.1210938,0.0002388183,0.0003803348],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.08174628,0.00001962384,0.914777,0.0002418861,0.0009762264,0.0003701661,0.00003836529,0.0007199245,0.001110572],"genre_scores_gemma":[0.9317078,0.000107041,0.06518055,0.0001108151,0.00005197922,0.000005871817,0.00003125631,0.00003707869,0.002767627],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.8499615,"threshold_uncertainty_score":0.9999049,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1660565028958392,"score_gpt":0.2201985164228235,"score_spread":0.05414201352698425,"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."}}