{"id":"W4414230282","doi":"10.1101/2025.09.11.675507","title":"Dual-LLM Adversarial Framework for Information Extraction from Research Literature","year":2025,"lang":"en","type":"preprint","venue":"bioRxiv (Cold Spring Harbor Laboratory)","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Prompt (Canada)","funders":"","keywords":"Adversarial system; Consistency (knowledge bases); Information extraction; Task (project management); Natural language; Process (computing); Data extraction; Task analysis","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":[0.005790181,0.001248676,0.001066898,0.00208499,0.0006006781,0.001260859,0.001982758,0.00176218,0.003004323],"category_scores_gemma":[0.01119532,0.0005388953,0.001011054,0.001235472,0.001408998,0.001870986,0.002752487,0.002138312,0.001048243],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001488138,"about_ca_system_score_gemma":0.001722985,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002592466,"about_ca_topic_score_gemma":0.003846415,"domain_scores_codex":[0.9977863,0.001051512,0.0001121145,0.0004277826,0.000475266,0.0001469064],"domain_scores_gemma":[0.9922031,0.005783496,0.0006363894,0.0006426713,0.0005455958,0.0001887199],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.0003209642,0.0001291762,0.002086714,0.0003166585,0.0001807205,0.0003369904,0.000221544,0.8206501,0.006307188,0.01980595,0.01022567,0.1394183],"study_design_scores_gemma":[0.000009015905,0.00002256902,0.0001306572,0.00001892032,0.00001157503,0.00003187341,0.000009916483,0.9877528,0.001424789,0.009526668,0.001052208,0.000009029232],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.02237809,0.001107886,0.9696604,0.001563848,0.0001050817,0.0001324113,0.0007345718,0.00229103,0.002026628],"genre_scores_gemma":[0.5756463,0.0006858738,0.4075729,0.001594053,0.0002834667,0.0005067602,0.003459509,0.0004217456,0.009829412],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.005790181,"threshold_uncertainty_score":0.03062171,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02110443440788621,"score_gpt":0.3097494355720354,"score_spread":0.2886450011641492,"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."}}