{"id":"W4388328907","doi":"10.48550/arxiv.2311.00913","title":"Self-Influence Guided Data Reweighting for Language Model Pre-training","year":2023,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Topic Modeling","field":"Computer Science","cited_by":2,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Natural Sciences and Engineering Research Council of Canada; Canadian Institute for Advanced Research","keywords":"Computer science; Novelty; Artificial intelligence; Context (archaeology); Task (project management); Relevance (law); Language model; Machine learning; Sample (material); Training set; Point (geometry); Stability (learning theory); Quality (philosophy); Natural language processing; Mathematics","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":[0.006003206,0.00193739,0.001833913,0.001406372,0.000970129,0.002011497,0.002827753,0.002220865,0.002768626],"category_scores_gemma":[0.03103354,0.00100723,0.0014502,0.001034458,0.001612653,0.003818194,0.003694574,0.005369117,0.002229148],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0008161211,"about_ca_system_score_gemma":0.00160074,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002157265,"about_ca_topic_score_gemma":0.006083444,"domain_scores_codex":[0.9967554,0.001419384,0.0002276036,0.0007839564,0.0006037286,0.0002099913],"domain_scores_gemma":[0.9880826,0.007024096,0.0005719938,0.002393532,0.00151533,0.0004124239],"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":[0.00109355,0.0006857196,0.009006938,0.0006468797,0.0005295777,0.0002970326,0.001028997,0.2980323,0.03775102,0.01475778,0.01337954,0.6227906],"study_design_scores_gemma":[0.00005903399,0.0001682811,0.0007368437,0.00004102575,0.000050146,0.00007737008,0.00006688235,0.9731771,0.0125761,0.01039053,0.002629642,0.00002697525],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.03570804,0.0008292208,0.9568722,0.0003855106,0.0001805133,0.0001501679,0.0002098013,0.004493878,0.001170692],"genre_scores_gemma":[0.49109,0.0004593207,0.4976164,0.0006567492,0.0004167378,0.0006419016,0.00221752,0.00221154,0.004689817],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.006003206,"threshold_uncertainty_score":0.03174835,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2694218530808317,"score_gpt":0.2625380589666247,"score_spread":0.006883794114206976,"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."}}