{"id":"W4404782398","doi":"10.18653/v1/2024.emnlp-main.606","title":"Stable Language Model Pre-training by Reducing Embedding Variability","year":2024,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"Kootenay Association for Science & Technology","funders":"Institute for Information and Communications Technology Promotion; Ministry of Science and ICT, South Korea; Korea Advanced Institute of Science and Technology","keywords":"Computer science; Training (meteorology); Embedding; Language model; Artificial intelligence; Natural language processing","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":[0.002666036,0.002866138,0.001229142,0.0009259846,0.0009033642,0.002057169,0.002203388,0.001679034,0.005580188],"category_scores_gemma":[0.01669428,0.001071783,0.001206293,0.0007994061,0.0008712594,0.004384035,0.002947765,0.005285389,0.004200273],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001083964,"about_ca_system_score_gemma":0.002641671,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.005476082,"about_ca_topic_score_gemma":0.01290237,"domain_scores_codex":[0.9984747,0.0004043787,0.000117221,0.0005087695,0.0002785989,0.0002162604],"domain_scores_gemma":[0.9952003,0.002436653,0.0002679982,0.0009151782,0.0009875858,0.0001922789],"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":[0.0007264233,0.0005418722,0.007695019,0.000399683,0.0003258476,0.0004182543,0.0006758529,0.3091508,0.08040499,0.007440934,0.01827691,0.5739434],"study_design_scores_gemma":[0.00004172618,0.000179932,0.001402425,0.0000383115,0.00006979781,0.0001102386,0.00009401311,0.9546521,0.03369811,0.007381127,0.002296595,0.00003566597],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.1009284,0.001188662,0.877589,0.0007146661,0.0002577071,0.0002506162,0.0005706853,0.01502433,0.003475931],"genre_scores_gemma":[0.725413,0.0004970123,0.259336,0.0008047438,0.000162439,0.0006125988,0.003338635,0.002291352,0.007544327],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.005580188,"threshold_uncertainty_score":0.01866758,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02390127958440878,"score_gpt":0.2959054770229754,"score_spread":0.2720041974385666,"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."}}