{"id":"W4402684242","doi":"10.18653/v1/2024.arabicnlp-1.23","title":"On the Utility of Pretraining Language Models on Synthetic Data","year":2024,"lang":"en","type":"article","venue":"","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Alliance de recherche numérique du Canada; Social Sciences and Humanities Research Council of Canada; Natural Sciences and Engineering Research Council of Canada; Canada Research Chairs","keywords":"Computer science; Natural language processing; Data modeling; Language model; Artificial intelligence; Data science; Database","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.005278997,0.0025538,0.0008298142,0.0009376674,0.0007314947,0.001592776,0.001893718,0.001935915,0.00281096],"category_scores_gemma":[0.02242182,0.0008318662,0.0009508604,0.0009269172,0.00111992,0.004008265,0.001672649,0.003077566,0.001807224],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00114987,"about_ca_system_score_gemma":0.001366647,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01639923,"about_ca_topic_score_gemma":0.02619207,"domain_scores_codex":[0.9974762,0.001480992,0.0001387371,0.0005163612,0.0002543503,0.000133361],"domain_scores_gemma":[0.9877751,0.008946494,0.0002813289,0.001615338,0.001215471,0.0001662692],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.0005198837,0.0006348134,0.008794424,0.0005059558,0.0004714102,0.0002554966,0.0002367894,0.7963132,0.01237296,0.002452323,0.01032088,0.1671219],"study_design_scores_gemma":[0.0000543224,0.0002960148,0.001947598,0.00007365047,0.00006050909,0.00009765163,0.0001469469,0.9748584,0.01685113,0.002052933,0.003511769,0.00004909296],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.6706671,0.00408513,0.2895099,0.00230665,0.0007003996,0.0005661411,0.006390669,0.01291294,0.01286112],"genre_scores_gemma":[0.851197,0.00109994,0.1233053,0.0007964757,0.00009807564,0.0004928532,0.01910498,0.0006174234,0.003287884],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01639923,"threshold_uncertainty_score":0.03260756,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.06481140776400941,"score_gpt":0.3232609385198948,"score_spread":0.2584495307558854,"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."}}