{"id":"W4407241185","doi":"10.1007/s10994-024-06670-4","title":"Schema-tune: noise-driven bias mitigation in transformer-based language models","year":2025,"lang":"en","type":"article","venue":"Machine Learning","topic":"Topic Modeling","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":false,"ca_institutions":"Toronto Metropolitan University","funders":"Natural Sciences and Engineering Research Council of Canada","keywords":"Computer science; Schema (genetic algorithms); Transformer; Artificial intelligence; Natural language processing; Machine learning; Speech recognition; Engineering; Electrical engineering","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":[0.0003383162,0.0001267053,0.0001602021,0.0002659285,0.00009708267,0.00008641922,0.000401627,0.00006697741,0.00001461387],"category_scores_gemma":[0.00007076745,0.0001282388,0.00006003261,0.0004931241,0.00001496293,0.0003222564,0.0000504464,0.0003581918,0.00001062228],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00006409177,"about_ca_system_score_gemma":0.00008142052,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0003598315,"about_ca_topic_score_gemma":0.0001689621,"domain_scores_codex":[0.9988691,0.0001177932,0.0002452044,0.0003368021,0.0001870582,0.00024407],"domain_scores_gemma":[0.9995098,0.00009373982,0.00005139382,0.0002796884,0.00002541232,0.00003995405],"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":[0.000006430073,0.00003377568,0.01734034,0.00005522046,0.00000741503,0.00001652805,0.001965757,0.8790865,0.005073826,0.01358184,0.0000110634,0.08282131],"study_design_scores_gemma":[0.000559248,0.00001667363,0.0007313705,0.00009451806,0.000003443886,9.376315e-7,0.00005183139,0.9940367,0.003206583,0.0008683797,0.0003089859,0.0001213267],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.2428226,0.000197946,0.7503588,0.001681598,0.0000766687,0.0001120605,5.174868e-7,0.0001858971,0.004563899],"genre_scores_gemma":[0.9564562,0.000004032177,0.04259426,0.0004117207,0.00002033548,0.00001580039,0.000009601928,0.000008965818,0.000479084],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.7136337,"threshold_uncertainty_score":0.5229428,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01802481198386612,"score_gpt":0.2643060661716281,"score_spread":0.246281254187762,"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."}}