{"id":"W6979294082","doi":"","title":"Challenging reaction prediction models to generalize to novel chemistry","year":2025,"lang":"en","type":"article","venue":"arXiv (Cornell University)","topic":"Machine Learning in Materials Science","field":"Materials Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Division of Chemistry; Natural Sciences and Engineering Research Council of Canada; Novo Nordisk; National Science Foundation","keywords":"Extrapolation; Benchmark (surveying); Predictive modelling; Core (optical fiber); Training set; Applicability domain; Deep learning","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":[0.004032664,0.00131395,0.0008272688,0.000728765,0.0004812787,0.001395729,0.001645777,0.001518275,0.002329089],"category_scores_gemma":[0.01188755,0.0005692051,0.001124545,0.000965703,0.000937274,0.00362449,0.001231317,0.004030869,0.001375057],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001629002,"about_ca_system_score_gemma":0.001682085,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01160937,"about_ca_topic_score_gemma":0.0149705,"domain_scores_codex":[0.9992697,0.0002379873,0.00004369417,0.0002641617,0.0001168254,0.00006761776],"domain_scores_gemma":[0.9944448,0.00342621,0.0003986696,0.0009357221,0.0006390623,0.0001555197],"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":[0.0001936294,0.0001497941,0.006375999,0.0001450876,0.00007206693,0.00007199762,0.00006387547,0.9425566,0.0009915447,0.00386841,0.004339828,0.04117114],"study_design_scores_gemma":[0.00001585556,0.0000412163,0.0003913047,0.00001244402,0.000007803616,0.0000121312,0.00001435808,0.9923202,0.0008954043,0.005542886,0.0007403215,0.000006027557],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.7105507,0.003355957,0.2543048,0.006701709,0.0004275506,0.000233025,0.005442972,0.005005517,0.01397766],"genre_scores_gemma":[0.9392483,0.0009740846,0.04694512,0.0008424816,0.0001361787,0.000227463,0.006367401,0.0002623289,0.004996658],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01160937,"threshold_uncertainty_score":0.02308363,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05885881642873665,"score_gpt":0.1960966733586483,"score_spread":0.1372378569299116,"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."}}