Reassessing High-Performing LLMs on Polish Medical Exams: True Competence or Bias-Driven Performance?
For researchers evaluating LLMs in medicine, this work demonstrates that standard MCQA benchmarks overestimate true clinical competence due to biases, though the contribution is incremental as it extends existing critique to a new language and domain.
The authors introduced an expanded Polish medical exam benchmark with over 15,000 questions and structural modifications to reduce MCQA artifacts, finding that evaluation design strongly affects LLM performance, with the best model dropping by 28.4 and 31 percentage points on English and Polish exams respectively under the harder setup.
Large language models (LLMs) in medicine are mainly evaluated using multiple-choice question answering (MCQA), which can overestimate real clinical ability due to guessing strategies and answer biases. To address these limitations, we introduce an expanded and more challenging benchmark based on Polish medical exams, adding over 15,000 questions, two new domains, and four structural modifications that reduce MCQA-specific artifacts and better test reasoning. We evaluate 21 LLMs and show that evaluation design strongly affects results. Under our harder setup, the best model (Qwen3.5-122B) drops by 28.4 and 31 pp on English and Polish exams, respectively. Despite low evidence of data contamination, standard MCQA scores do not reliably reflect true medical competence. To facilitate further research, we make our benchmark publicly available.