{"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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.007732051,0.002490298,0.001807803,0.001153233,0.0009512914,0.002677694,0.002969412,0.002597935,0.01320821],"category_scores_gemma":[0.02089291,0.0008234787,0.0009436826,0.001114491,0.001563658,0.005820203,0.003260248,0.005548948,0.004864404],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002295093,"about_ca_system_score_gemma":0.001717787,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.005430872,"about_ca_topic_score_gemma":0.007982687,"domain_scores_codex":[0.9973639,0.001288756,0.0001364376,0.0004378023,0.0005409461,0.0002321157],"domain_scores_gemma":[0.9913421,0.006807243,0.0001983568,0.0009316665,0.0005508445,0.0001699064],"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":[0.0004315766,0.000188635,0.00114233,0.0003235209,0.0001587844,0.0001979108,0.000107113,0.6622263,0.001490246,0.09334208,0.02986496,0.2105265],"study_design_scores_gemma":[0.00002118522,0.00003766304,0.0001327241,0.00002796033,0.000009741106,0.00003566888,0.00001288572,0.9488022,0.0004956172,0.04823252,0.002182876,0.000008968649],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.02650611,0.004089254,0.945612,0.004141415,0.0005925613,0.0001425755,0.0009807198,0.004882189,0.01305319],"genre_scores_gemma":[0.5868233,0.002937603,0.3482168,0.003383072,0.001167063,0.0008386393,0.004928016,0.00217981,0.04952566],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.01320821,"threshold_uncertainty_score":0.04418588,"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."}}