# It is Too Many Options: Pitfalls of Multiple-Choice Questions in Generative AI and Medical Education

arXiv: https://arxiv.org/abs/2503.13508 · HTML: https://mlsanity.com/paper/2503.13508/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- Builds on 12 of its 22 references that are in the map.
- First posted on arXiv 13 Mar 2025.

## Builds on (most cited first, 12 of 12)

- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [On the Opportunities and Risks of Foundation Models](https://mlsanity.com/paper/2108.07258/) (2021)
- [Large Language Models Encode Clinical Knowledge](https://mlsanity.com/paper/2212.13138/) (2022)
- [Capabilities of GPT-4 on Medical Challenge Problems](https://mlsanity.com/paper/2303.13375/) (2023)
- [Can Generalist Foundation Models Outcompete Special-Purpose Tuning? Case Study in Medicine](https://mlsanity.com/paper/2311.16452/) (2023)
- [Can multiple-choice questions really be useful in detecting the abilities of LLMs?](https://mlsanity.com/paper/2403.17752/) (2024)
- [LongHealth: A Question Answering Benchmark with Long Clinical Documents](https://mlsanity.com/paper/2401.14490/) (2024)
- [MedBench: A Large-Scale Chinese Benchmark for Evaluating Medical Large Language Models](https://mlsanity.com/paper/2312.12806/) (2023)
- [Rad-ReStruct: A Novel VQA Benchmark and Method for Structured Radiology Reporting](https://mlsanity.com/paper/2307.05766/) (2023)
- [Pattern Recognition or Medical Knowledge? The Problem with Multiple-Choice Questions in Medicine](https://mlsanity.com/paper/2406.02394/) (2024)
- [Is Your Large Language Model Knowledgeable or a Choices-Only Cheater?](https://mlsanity.com/paper/2407.01992/) (2024)
- [Assessing The Potential Of Mid-Sized Language Models For Clinical QA](https://mlsanity.com/paper/2404.15894/) (2024)

## Built on by (most cited first, 7 of 7)

- [Movie Facts and Fibs (MF2): A Benchmark for Long Movie Understanding](https://mlsanity.com/paper/2506.06275/) (2025)
- [RubricsTree: Scalable and Evolving Open-Ended Evaluation of Personal Health Agents across Health Memory and Medical Skills](https://mlsanity.com/paper/2606.18203/) (2026)
- [Improving the Data-efficiency of Reinforcement Learning by Warm-starting with LLM](https://mlsanity.com/paper/2505.10861/) (2025)
- [On the Robustness of Answer Formats in Medical Reasoning Models](https://mlsanity.com/paper/2509.20866/) (2025)
- [KoSimpleQA: A Korean Factuality Benchmark with an Analysis of Reasoning LLMs](https://mlsanity.com/paper/2510.18368/) (2025)
- [Healthier LLMs: Retrieval-Augmented Generation for Public Health Question Answering](https://mlsanity.com/paper/2607.06641/) (2026)
- [Synthetic Hospital: An Open, Verifiable, Physician-Validated Longitudinal EHR Benchmark](https://mlsanity.com/paper/2609.30027/) (2026)

## Abstract (arXiv)

The performance of Large Language Models (LLMs) on multiple-choice question (MCQ) benchmarks is frequently cited as proof of their medical capabilities. We hypothesized that LLM performance on medical MCQs may in part be illusory and driven by factors beyond medical content knowledge and reasoning capabilities. To assess this, we created a novel benchmark of free-response questions with paired MCQs (FreeMedQA). Using this benchmark, we evaluated three state-of-the-art LLMs (GPT-4o, GPT-3.5, and LLama-3-70B-instruct) and found an average absolute deterioration of 39.43% in performance on free-response questions relative to multiple-choice (p = 1.3 * 10-5) which was greater than the human performance decline of 22.29%. To isolate the role of the MCQ format on performance, we performed a masking study, iteratively masking out parts of the question stem. At 100% masking, the average LLM multiple-choice performance was 6.70% greater than random chance (p = 0.002) with one LLM (GPT-4o) obtaining an accuracy of 37.34%. Notably, for all LLMs the free-response performance was near zero. Our results highlight the shortcomings in medical MCQ benchmarks for overestimating the capabilities of LLMs in medicine, and, broadly, the potential for improving both human and machine assessments using LLM-evaluated free-response questions.