{"id":"W4408173301","doi":"10.26434/chemrxiv-2025-8z6h2","title":"MERMaid: Universal multimodal mining of chemical reactions from PDFs using vision-language models","year":2025,"lang":"en","type":"preprint","venue":"ChemRxiv","topic":"Advanced Text Analysis Techniques","field":"Computer Science","cited_by":3,"is_retracted":false,"has_abstract":true,"ca_institutions":"Vector Institute; Canadian Institute for Advanced Research; University of Toronto","funders":"Canada First Research Excellence Fund; Canadian Institute for Advanced Research","keywords":"Computer science; Artificial intelligence","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":[0.001410358,0.001660963,0.0007952857,0.00268861,0.0007381454,0.002616288,0.002412057,0.001579004,0.00816145],"category_scores_gemma":[0.006457157,0.0006925426,0.001556202,0.001280237,0.0006364748,0.003265661,0.003476038,0.001798647,0.00482051],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0008904304,"about_ca_system_score_gemma":0.001495901,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003560912,"about_ca_topic_score_gemma":0.007003776,"domain_scores_codex":[0.9991646,0.0001385124,0.00004984531,0.0003397555,0.0002498734,0.00005749188],"domain_scores_gemma":[0.9985803,0.0007243589,0.0001166544,0.0002844736,0.0002241316,0.00007000179],"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":[0.0008297698,0.0002940276,0.002825018,0.001488351,0.0003063767,0.0006417232,0.0008630602,0.03020214,0.06255092,0.0196915,0.07583051,0.8044766],"study_design_scores_gemma":[0.00009754256,0.0001394876,0.001108599,0.0001234517,0.00007988845,0.0004470914,0.0003068536,0.8086317,0.0719439,0.05685798,0.0601542,0.0001092991],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"software","genre_scores_codex":[0.01047494,0.0007128593,0.8903714,0.0003462612,0.0001633746,0.0002371951,0.006601264,0.08770954,0.003383152],"genre_scores_gemma":[0.1023001,0.0005228141,0.8768252,0.0004259826,0.00006354368,0.0005310808,0.01256174,0.002376096,0.004393488],"genre_candidate":"software","genre_consensus":null,"teacher_disagreement_score":0.00816145,"threshold_uncertainty_score":0.0273028,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02439863093086714,"score_gpt":0.3166167865368164,"score_spread":0.2922181556059493,"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."}}