{"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":"metacan-v3-hybrid-931329e0061c","candidate_categories":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.1459157,0.001875887,0.001629851,0.01463591,0.002632358,0.007305385,0.003658397,0.002960693,0.003510344],"category_scores_gemma":[0.4168662,0.001449095,0.004182865,0.01254653,0.001516235,0.00790455,0.007950903,0.002890356,0.002235482],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.005253008,"about_ca_system_score_gemma":0.02230133,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.009301137,"about_ca_topic_score_gemma":0.03103614,"domain_scores_codex":[0.8376863,0.1214095,0.02398834,0.005210578,0.01075612,0.0009491426],"domain_scores_gemma":[0.3495971,0.5594947,0.02420487,0.03510545,0.03005064,0.001547203],"domain_codex":null,"domain_gemma":"methods","domain_candidate":"methods","domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"qualitative","study_design_scores_codex":[0.001233229,0.0005319326,0.01644078,0.07687446,0.001449829,0.00252322,0.02420903,0.01335609,0.009068253,0.01339855,0.03517033,0.8057443],"study_design_scores_gemma":[0.001100025,0.001437779,0.02656251,0.093274,0.006836943,0.005268539,0.02212423,0.1816469,0.03678589,0.07356855,0.5500176,0.001377052],"study_design_candidate":"qualitative","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.1047261,0.03470704,0.7505367,0.02654315,0.001158243,0.01637432,0.01950336,0.02679081,0.0196603],"genre_scores_gemma":[0.140074,0.006610583,0.8333712,0.002270247,0.0001552952,0.007478823,0.00707949,0.001448029,0.001512318],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.8540843,"threshold_uncertainty_score":0.7716851,"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."}}