{"id":"W4412599472","doi":"10.1016/j.ijmedinf.2025.106048","title":"Leveraging open-source large language models (LLMs) in scoping reviews: a case study on disability and AI applications","year":2025,"lang":"en","type":"article","venue":"International Journal of Medical Informatics","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":7,"is_retracted":false,"has_abstract":true,"ca_institutions":"Université de Montréal; Centre Intégré Universitaire de Santé et de Services Sociaux du Centre-Sud-de-l'Île-de-Montréal","funders":"Université de Montréal","keywords":"Open source; Computer science; Data extraction; Data science; Data source; Natural language processing; Data mining; MEDLINE; Political science","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":[],"consensus_categories":[],"category_scores_codex":[0.002597862,0.00008414632,0.0002865214,0.0001964985,0.00006046025,0.00008558029,0.0003140319,0.00007276781,0.00004819901],"category_scores_gemma":[0.0007712056,0.00006569622,0.00004355154,0.0001839103,0.00005953391,0.0003491188,0.0001504635,0.0005190538,0.000004965303],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002039149,"about_ca_system_score_gemma":0.0003745646,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0004555629,"about_ca_topic_score_gemma":0.0002237246,"domain_scores_codex":[0.997835,0.00007853393,0.001282324,0.00007409108,0.0006070997,0.0001229942],"domain_scores_gemma":[0.9988719,0.0003080976,0.0002682794,0.0001493562,0.0002530588,0.0001492866],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"qualitative","study_design_scores_codex":[0.000271727,0.003351024,0.1019873,0.0008317771,0.0001706656,0.0006740321,0.08677758,0.0003730945,0.000004369517,0.002227409,0.001390747,0.8019403],"study_design_scores_gemma":[0.004964791,0.001141914,0.004813133,0.0340175,0.0002609642,0.008688187,0.6872066,0.2353617,0.0003095658,0.00634235,0.01634662,0.000546752],"study_design_candidate":"design_other","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9194435,0.0005199461,0.06777005,0.01042266,0.0003294452,0.0009601252,0.000001991464,0.000007686132,0.0005446259],"genre_scores_gemma":[0.9934142,0.0004034678,0.0005806754,0.005343348,0.0001801904,0.00003936243,0.000003926111,0.000004207336,0.00003060963],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.8013936,"threshold_uncertainty_score":0.2679014,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1687977195233445,"score_gpt":0.5295397091899532,"score_spread":0.3607419896666087,"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."}}