{"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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0004138921,0.0001493714,0.0001585913,0.0001404661,0.0002423778,0.00007581845,0.0004726527,0.00008015687,0.0002115322],"category_scores_gemma":[0.00007702618,0.0001724791,0.00004121583,0.0006527831,0.00003976921,0.0003702384,0.0002562336,0.00009659868,0.0002623538],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002191089,"about_ca_system_score_gemma":0.00005317663,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0002556292,"about_ca_topic_score_gemma":0.00001294871,"domain_scores_codex":[0.9987152,0.0000433338,0.0001571287,0.0006723529,0.00009665618,0.0003153832],"domain_scores_gemma":[0.9992152,0.00003422669,0.00006298169,0.0004362519,0.00009538202,0.0001559982],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"bench_or_experimental","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00002973544,0.00002375901,0.00007275939,0.00001434541,0.000002191209,0.000003329418,0.00008131717,0.4483885,0.5448518,0.006362678,0.000141623,0.00002799739],"study_design_scores_gemma":[0.0006851206,0.00006187916,0.001709167,0.0001374934,0.00003996407,0.000006653849,0.0002677095,0.7768521,0.2130972,0.003474215,0.003251863,0.0004166182],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.6502474,0.0000030839,0.3446387,0.0002067193,0.0004247546,0.0001292647,0.00001208476,0.000177698,0.004160282],"genre_scores_gemma":[0.9922343,0.000006557198,0.002556289,0.0002665195,0.00007868664,0.000002365837,0.000003740211,0.00001051251,0.004841037],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.3420824,"threshold_uncertainty_score":0.7033494,"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."}}