{"id":"W4416230956","doi":"10.26434/chemrxiv-2025-6dj8j","title":"Bringing ML to the real world: rewards are all we need","year":2025,"lang":"","type":"article","venue":"ChemRxiv","topic":"Machine Learning in Materials Science","field":"Materials Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Artificial Intelligence in Medicine (Canada)","funders":"","keywords":"Operationalization; Extant taxon; Function (biology); Key (lock); Measure (data warehouse); Human intelligence","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.006438453,0.001367657,0.001415412,0.001152985,0.001616439,0.007791094,0.002620847,0.003081667,0.01029034],"category_scores_gemma":[0.03947546,0.0008005836,0.0007632941,0.001184578,0.009760684,0.01524061,0.004761317,0.006481265,0.002988158],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002799095,"about_ca_system_score_gemma":0.003194318,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00157895,"about_ca_topic_score_gemma":0.002233309,"domain_scores_codex":[0.9951656,0.00256778,0.000265034,0.0007294877,0.0009725275,0.0002995679],"domain_scores_gemma":[0.9790794,0.014967,0.001330328,0.002400168,0.001335665,0.0008874288],"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":[0.0001216202,0.00006414438,0.001289647,0.000499023,0.00006350147,0.0001114259,0.0005775947,0.03495935,0.00100881,0.8324019,0.01249775,0.1164054],"study_design_scores_gemma":[0.00002601235,0.00003983727,0.0002040323,0.0001838394,0.00001958668,0.00006522246,0.0001133902,0.03921086,0.0006065362,0.9372611,0.02222965,0.00003982979],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.008446023,0.008645361,0.9097224,0.03845279,0.0007229386,0.000101735,0.0002792746,0.001334979,0.03229444],"genre_scores_gemma":[0.5443699,0.008813566,0.4226784,0.007957739,0.001352285,0.0003998387,0.0004733526,0.001414192,0.01254079],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.01029034,"threshold_uncertainty_score":0.03442454,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01820846254988943,"score_gpt":0.3046869061053663,"score_spread":0.2864784435554769,"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."}}