{"id":"W3176692111","doi":"10.18653/v1/2021.findings-acl.119","title":"LICHEE: Improving Language Model Pre-training with Multi-grained Tokenization","year":2021,"lang":"en","type":"preprint","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":5,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Alberta","funders":"","keywords":"Computer science; Lexical analysis; Language model; Benchmark (surveying); Artificial intelligence; Natural language processing; Inference; Variety (cybernetics); Natural language understanding; Representation (politics); Natural language; Machine learning","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"],"consensus_categories":[],"category_scores_codex":[0.0002929792,0.0003087962,0.0003213901,0.0001261099,0.00009123203,0.0006085067,0.001069775,0.0002344728,0.00001315907],"category_scores_gemma":[0.00008243391,0.0002718895,0.00008401067,0.000202804,0.00001902029,0.0004111767,0.001917946,0.0005190353,0.000002407329],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00009183007,"about_ca_system_score_gemma":0.0006141505,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0003905954,"about_ca_topic_score_gemma":0.0002753336,"domain_scores_codex":[0.9978448,0.00005613595,0.0003387244,0.001016487,0.0003785937,0.0003653217],"domain_scores_gemma":[0.9981411,0.00003413829,0.0001924817,0.001327859,0.000191773,0.000112625],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.000003983485,0.00005272982,0.0001555846,0.0001943385,0.00004889388,0.00004585257,0.03921345,0.8893139,0.003283044,0.002967849,0.00001316907,0.06470724],"study_design_scores_gemma":[0.0002886061,0.0000117022,0.00009879472,0.0001269791,0.00001435737,0.00001015319,0.0005708443,0.9973983,0.001007688,0.000113798,0.00000296649,0.0003558664],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.07454801,0.0001387627,0.9229177,0.0002364881,0.0002229119,0.0003181273,0.000002154883,0.0006203571,0.0009954678],"genre_scores_gemma":[0.4887787,0.000003218832,0.509748,0.000197335,0.00006018455,0.00002986363,0.00002178641,0.00001896544,0.001141956],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.4142307,"threshold_uncertainty_score":0.9999734,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03371922272302266,"score_gpt":0.2704040480036245,"score_spread":0.2366848252806018,"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."}}