{"id":"W6910249080","doi":"10.48448/4a80-vz17","title":"LICHEE: Improving Language Model Pre-training with Multi-grained Tokenization","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 Alberta","funders":"","keywords":"Language model; Variety (cybernetics); Natural language understanding; Natural language; Benchmark (surveying); Representation (politics); Language identification; Inference","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":[0.001922676,0.002349781,0.001778494,0.001617201,0.0008937887,0.001267823,0.003203208,0.001576815,0.007124363],"category_scores_gemma":[0.006030423,0.000994875,0.001471805,0.00135472,0.0008495472,0.004989631,0.003204178,0.004503935,0.006016779],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.000966089,"about_ca_system_score_gemma":0.003100639,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01549457,"about_ca_topic_score_gemma":0.03314811,"domain_scores_codex":[0.9984925,0.0004610404,0.00009587363,0.0005243999,0.0002298109,0.0001964623],"domain_scores_gemma":[0.997624,0.001107092,0.00009125527,0.0006386348,0.000413638,0.0001254313],"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":[0.0005158623,0.0005513875,0.002821186,0.0003487986,0.0002772199,0.0004215257,0.000358846,0.1532758,0.02082896,0.005483769,0.04313001,0.7719866],"study_design_scores_gemma":[0.00006621355,0.0001430717,0.0005040253,0.00002727571,0.00004703889,0.000124349,0.00009661138,0.9717738,0.0131156,0.005463405,0.008575504,0.00006302524],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.03258209,0.001530071,0.9156922,0.0004973733,0.0004725089,0.0002258665,0.001359995,0.04386264,0.003777194],"genre_scores_gemma":[0.338724,0.0007246007,0.6223899,0.001403751,0.0002081559,0.0007184361,0.01525984,0.00360921,0.0169621],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.01549457,"threshold_uncertainty_score":0.03080881,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03528229445213348,"score_gpt":0.3160258786409297,"score_spread":0.2807435841887962,"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."}}