{"id":"W3092435621","doi":"10.18653/v1/2020.emnlp-main.403","title":"On Losses for Modern Language Models","year":2020,"lang":"en","type":"preprint","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":2,"is_retracted":false,"has_abstract":true,"ca_institutions":"St. Michael's Hospital; Vector Institute; University of Toronto","funders":"Canadian Institute for Advanced Research","keywords":"Computer science; Sentence; Benchmark (surveying); Task (project management); Baseline (sea); Natural language processing; Set (abstract data type); Context (archaeology); Artificial intelligence; Training set; Machine learning; Language model; Speech recognition; Programming language","routes":{"ca_aff":true,"ca_fund":true,"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":[],"consensus_categories":[],"category_scores_codex":[0.00009692828,0.0001944963,0.0002329808,0.00005720293,0.00003789412,0.000199586,0.001392074,0.0001447328,0.000008763213],"category_scores_gemma":[0.00003753704,0.0001740248,0.0001364198,0.00004173237,0.0000080715,0.0001256062,0.00128864,0.0002661367,0.00002295998],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00003818488,"about_ca_system_score_gemma":0.0001106054,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00005946638,"about_ca_topic_score_gemma":0.000007376729,"domain_scores_codex":[0.9985815,0.0000179014,0.0001976064,0.0007648413,0.0002274985,0.0002106316],"domain_scores_gemma":[0.9986991,0.0001016292,0.00006949401,0.0009964021,0.00004804027,0.00008533065],"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":[0.000005443401,0.00002359964,4.31944e-7,0.0001257364,0.00002208856,0.00001188797,0.001747492,0.3952659,0.00009086048,0.5659719,0.001325444,0.03540923],"study_design_scores_gemma":[0.00006987539,0.00001392928,3.437752e-7,0.00002189111,0.000002706357,5.718128e-7,0.000006923073,0.5884508,0.000313456,0.4109589,0.00004326948,0.0001173084],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.002107178,0.0001264317,0.9816288,0.002783351,0.0004451975,0.0004066644,0.00001358446,0.0004645369,0.01202422],"genre_scores_gemma":[0.6207591,0.00000552008,0.3764652,0.00162023,0.0001639629,0.00007161759,0.000009364417,0.00001612451,0.0008888948],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.6186519,"threshold_uncertainty_score":0.7096526,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.07732662353992366,"score_gpt":0.3006394215511403,"score_spread":0.2233127980112167,"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."}}