{"id":"W4413095517","doi":"10.1016/j.matt.2025.102331","title":"MERMaid: Universal multimodal mining of chemical reactions from PDFs using vision-language models","year":2025,"lang":"en","type":"article","venue":"Matter","topic":"Machine Learning in Materials Science","field":"Materials Science","cited_by":6,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"University of Toronto; Nanyang Technological University; Ministry of Education - Singapore; Canada First Research Excellence Fund; Canada Research Chairs; Canadian Institute for Advanced Research","keywords":"Computer science; Artificial intelligence; Natural language processing","routes":{"ca_aff":true,"ca_fund":true,"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":[0.001195781,0.001538994,0.0007865342,0.001929395,0.0006720248,0.00227465,0.002418797,0.001421193,0.007423216],"category_scores_gemma":[0.005733206,0.0006587659,0.001541476,0.0009295577,0.0006565787,0.002973508,0.003320087,0.001713828,0.003566755],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0008792196,"about_ca_system_score_gemma":0.001356793,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003634931,"about_ca_topic_score_gemma":0.007356811,"domain_scores_codex":[0.9993114,0.0001071644,0.00003754837,0.0002807865,0.0002079897,0.00005511541],"domain_scores_gemma":[0.9989536,0.0005315923,0.00007939083,0.0002063801,0.0001753265,0.00005372026],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0007114289,0.0003270612,0.002760952,0.001120841,0.0002481882,0.0006615405,0.0006867659,0.05089544,0.05971465,0.0233472,0.04748488,0.812041],"study_design_scores_gemma":[0.00006009793,0.0001057241,0.0006897742,0.00006970124,0.00004440502,0.0002897748,0.0001841483,0.878958,0.05026001,0.04296143,0.02630158,0.00007545247],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.01092868,0.0004506127,0.9236699,0.0002231327,0.0001196291,0.0001963099,0.002861702,0.05852341,0.00302675],"genre_scores_gemma":[0.1355351,0.0003732343,0.8509309,0.0003727948,0.00004150021,0.0004761629,0.00672978,0.001910854,0.003629643],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.007423216,"threshold_uncertainty_score":0.02483314,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01025074582778698,"score_gpt":0.2856658686287501,"score_spread":0.2754151228009631,"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."}}