{"id":"W6891824703","doi":"10.48448/2an9-z413","title":"LICHEE: Improving Language Model Pre-training with Multi-grained Tokenization","year":2021,"lang":"en","type":"other","venue":"Open MIND","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); Inference; Language identification","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":[0.001886104,0.002426041,0.001661796,0.001616711,0.0009209324,0.001337936,0.003514652,0.001701489,0.008538904],"category_scores_gemma":[0.006152338,0.0009924864,0.001484568,0.001434319,0.0008819767,0.005093907,0.003167046,0.004525988,0.006920829],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001021276,"about_ca_system_score_gemma":0.002953944,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01677092,"about_ca_topic_score_gemma":0.03797993,"domain_scores_codex":[0.9985831,0.0003968228,0.00008863184,0.0005210842,0.0002238193,0.0001865479],"domain_scores_gemma":[0.9977635,0.00105748,0.00008050528,0.000593829,0.000389367,0.000115232],"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":[0.0004956313,0.0005352236,0.002678642,0.0003674026,0.0002883163,0.0004287661,0.0003652767,0.1439797,0.01719363,0.006802066,0.06028608,0.7665794],"study_design_scores_gemma":[0.00006818362,0.0001341261,0.0004766619,0.0000305195,0.00004724122,0.0001312615,0.00009379158,0.9694473,0.0130303,0.006526754,0.0099511,0.00006273193],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.03853145,0.001753781,0.8918007,0.0005428224,0.0005751887,0.0002509244,0.00198485,0.05873634,0.005824012],"genre_scores_gemma":[0.3253792,0.0007099916,0.6257211,0.001454919,0.0001998168,0.0007270025,0.02024899,0.004438796,0.02112019],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.01677092,"threshold_uncertainty_score":0.03334659,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.0519361819426249,"score_gpt":0.3339995148840413,"score_spread":0.2820633329414164,"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."}}