{"id":"W4410022944","doi":"10.1088/2632-2153/add3bb","title":"LLM4Mat-bench: benchmarking large language models for materials property prediction","year":2025,"lang":"en","type":"article","venue":"Machine Learning Science and Technology","topic":"Machine Learning in Materials Science","field":"Materials Science","cited_by":13,"is_retracted":false,"has_abstract":true,"ca_institutions":"Schwartz/Reisman Emergency Medicine Institute; Vector Institute; Structural Genomics Consortium; University of Toronto","funders":"Office of Advanced Cyberinfrastructure; National Science Foundation","keywords":"Benchmarking; Property (philosophy); Computer science; Natural language processing; Business; Philosophy","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":["sts"],"consensus_categories":[],"category_scores_codex":[0.004934681,0.0002757954,0.0003949181,0.0009103853,0.001622939,0.0004812069,0.0009896313,0.000198561,0.0002249266],"category_scores_gemma":[0.001824266,0.0001984069,0.00002961119,0.001597127,0.001086586,0.0007051547,0.0007328772,0.0003190526,0.00002426959],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.000107918,"about_ca_system_score_gemma":0.0002559245,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0002125752,"about_ca_topic_score_gemma":0.00003022498,"domain_scores_codex":[0.9969782,0.0001238161,0.0004664271,0.001041357,0.0004859707,0.0009041762],"domain_scores_gemma":[0.9987484,0.0001169827,0.0002285667,0.0004853093,0.0003363917,0.00008438173],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"bench_or_experimental","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.00002802644,0.00003861727,0.00268347,0.00009237203,0.00000369949,0.000003178822,0.0003396538,0.001006675,0.9623359,0.02747501,0.0001581179,0.005835308],"study_design_scores_gemma":[0.001122951,0.0005127015,0.001055036,0.0001936281,0.00004203129,0.0000578367,0.0006106168,0.4282701,0.5414231,0.01220924,0.01398686,0.0005158768],"study_design_candidate":"bench_or_experimental","study_design_consensus":"bench_or_experimental","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9555829,0.0003133468,0.03577893,0.003310133,0.001243928,0.0007722676,0.00006439527,0.0009927773,0.001941297],"genre_scores_gemma":[0.984908,0.00002683445,0.01309844,0.0002950292,0.0000806263,0.0001666747,0.00001825156,0.00002033185,0.001385818],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.4272634,"threshold_uncertainty_score":0.9996768,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.007163373249796719,"score_gpt":0.2604273323180107,"score_spread":0.253263959068214,"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."}}