{"id":"W4404764563","doi":"10.1038/s41562-024-02046-9","title":"Large language models surpass human experts in predicting neuroscience results","year":2024,"lang":"en","type":"article","venue":"Nature Human Behaviour","topic":"Topic Modeling","field":"Computer Science","cited_by":111,"is_retracted":false,"has_abstract":true,"ca_institutions":"Institut Universitaire de Gériatrie de Montréal","funders":"National Eye Institute; Economic and Social Research Council; Research Councils UK; Royal Society","keywords":"Cognitive science; Psychology; Neuroscience","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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0008138698,0.0002220213,0.0001923186,0.0003517741,0.0002623614,0.0004819967,0.00128428,0.0002886211,0.000006460398],"category_scores_gemma":[0.00005426828,0.0002112995,0.00008396388,0.0005668855,0.00003222503,0.001033158,0.0004784864,0.001105917,0.000004887436],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001127501,"about_ca_system_score_gemma":0.00005626788,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001532107,"about_ca_topic_score_gemma":0.0004179715,"domain_scores_codex":[0.9973108,0.00008972242,0.0004212873,0.001019472,0.0005886201,0.000570163],"domain_scores_gemma":[0.9988863,0.00004471466,0.00006410585,0.0008479605,0.00004525126,0.0001116524],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00002112618,0.0009387374,0.03580713,0.0002218491,0.00001786615,0.007910031,0.1216123,0.005677221,0.09778042,0.7102785,0.009922543,0.009812206],"study_design_scores_gemma":[0.001996317,0.0002038022,0.04811908,0.0007792332,0.00002445269,0.0001444071,0.001096842,0.930577,0.00377501,0.01001396,0.00194358,0.001326297],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9681684,0.001117841,0.0253347,0.0004132992,0.001241237,0.000250651,0.00002529255,0.0007752129,0.002673327],"genre_scores_gemma":[0.9971087,0.000002950424,0.001178678,0.0003205361,0.0002083875,0.00002143493,0.00001298162,0.00002469259,0.001121631],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.9248998,"threshold_uncertainty_score":0.8616545,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02928450250818994,"score_gpt":0.3250620535175684,"score_spread":0.2957775510093785,"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."}}