{"id":"W4410400158","doi":"10.7759/cureus.84165","title":"Assessing Large Language Models for Medical Question Answering in Portuguese: Open-Source Versus Closed-Source Approaches","year":2025,"lang":"en","type":"article","venue":"Cureus","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":2,"is_retracted":false,"has_abstract":true,"ca_institutions":"Artificial Intelligence in Medicine (Canada)","funders":"","keywords":"Medicine; Open source; Portuguese; Source model; Natural language processing; Linguistics; Programming language","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":[0.02545453,0.002434673,0.001034851,0.003128243,0.0007506297,0.003874542,0.002468069,0.001989441,0.003426939],"category_scores_gemma":[0.1410364,0.000664371,0.001962731,0.002183746,0.0007640109,0.004293085,0.004369409,0.002591472,0.002202921],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002395472,"about_ca_system_score_gemma":0.002882587,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.008282084,"about_ca_topic_score_gemma":0.009914128,"domain_scores_codex":[0.9796869,0.01523544,0.001099021,0.002346274,0.001362446,0.0002699543],"domain_scores_gemma":[0.8584599,0.12591,0.002561272,0.006376235,0.005608561,0.001083984],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.003684542,0.00266483,0.05604751,0.004507702,0.002109699,0.0009345811,0.005686969,0.2203203,0.006948049,0.0093254,0.0353123,0.6524581],"study_design_scores_gemma":[0.000575477,0.0008464593,0.01755072,0.0008944264,0.0004662731,0.000356516,0.001672273,0.9178041,0.007562303,0.02938051,0.02266295,0.0002280647],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.4104781,0.005463767,0.5046301,0.005205221,0.0008789443,0.002452503,0.02020201,0.03757067,0.01311871],"genre_scores_gemma":[0.7478234,0.0007138947,0.2249028,0.0008171109,0.0001834815,0.001604571,0.02138346,0.0009849459,0.001586448],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.02545453,"threshold_uncertainty_score":0.134618,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2472598250764961,"score_gpt":0.4845967357826596,"score_spread":0.2373369107061635,"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."}}