Assessing Large Language Models for Medical Question Answering in Portuguese: Open-Source Versus Closed-Source Approaches
Bibliographic record
Abstract
Large language models (LLMs) show promise in medical knowledge assessment. This study benchmarked a closed-source (GPT-4o, OpenAI, San Francisco, CA) and an open-source (LLaMA 3.1 405B, Meta AI, Menlo Park, CA) LLM on 148 multiple-choice questions from the 2023 Portuguese National Residency Access Examination across five clinical domains. Using five distinct prompting strategies, models provided single-best-answer predictions. GPT-4o consistently outperformed LLaMA 3.1 by 7-11% accuracy across all prompts. Chain-of-thought prompting yielded the highest numerical accuracy for GPT-4o, though this improvement was not statistically significant over simpler prompts in post-hoc analyses, while offering minimal benefit when applied to LLaMA 3.1. Both models performed best in pediatrics and less accurately in surgery and psychiatry questions. Bias assessment indicated GPT-4o aligned well with correct answer distributions, unlike LLaMA 3.1, which showed prompt-dependent skew. Closed-source models currently demonstrate higher accuracy on Portuguese medical questions, likely due to extensive training. However, open-source models remain valuable for data control, though domain-focused fine-tuning may be needed for optimal performance in high-stakes applications.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.025 | 0.141 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.003 | 0.002 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.004 | 0.004 |
| Open science | 0.002 | 0.004 |
| Research integrity | 0.002 | 0.003 |
| Insufficient payload (model declined to judge) | 0.003 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".