{"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":"codex-gemma-dda1882f352a","candidate_categories":["insufficient_payload"],"consensus_categories":[],"category_scores_codex":[0.0002896742,0.000135455,0.0002288209,0.000111668,0.00008776849,0.00008039705,0.0003876433,0.0000781258,0.003712872],"category_scores_gemma":[0.00005817524,0.0001267321,0.00004949321,0.0001722112,0.0001705571,0.0003405664,0.0002567872,0.00009991841,0.0001663589],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0000611459,"about_ca_system_score_gemma":0.00005189874,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002072911,"about_ca_topic_score_gemma":0.000004525031,"domain_scores_codex":[0.9987255,0.000111494,0.0003212597,0.0003600161,0.000238377,0.0002433956],"domain_scores_gemma":[0.9992276,0.0001568771,0.0001446837,0.0003747964,0.00004755116,0.00004852831],"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":[0.00002424841,0.00003039063,0.0007771599,0.00002396938,0.000004698843,0.000004533059,0.0007160603,0.009876583,0.9878341,0.00008621725,0.0005428625,0.00007920412],"study_design_scores_gemma":[0.000404382,0.00001179896,0.001551049,0.0001859665,0.00003367995,0.000004886374,0.0004083143,0.2988931,0.6975175,0.0005981046,0.000186394,0.0002048546],"study_design_candidate":"bench_or_experimental","study_design_consensus":"bench_or_experimental","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9806855,0.00001974231,0.01545259,0.0002714713,0.0005343988,0.00008578546,0.00004411789,0.00006108863,0.002845298],"genre_scores_gemma":[0.9453511,3.794209e-7,0.05388859,0.0002879137,0.00005512107,0.000002823176,0.00000845384,0.00001217252,0.0003934158],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.2903166,"threshold_uncertainty_score":0.9971979,"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."}}