# MedDialBench: Benchmarking LLM Diagnostic Robustness under Parametric Adversarial Patient Behaviors

arXiv: https://arxiv.org/abs/2604.06846 · HTML: https://mlsanity.com/paper/2604.06846/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Builds on 16 of its 26 references that are in the map.
- First posted on arXiv 8 Apr 2026.

## Builds on (most cited first, 16 of 16)

- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [What Disease does this Patient Have? A Large-scale Open Domain Question Answering Dataset from Medical Exams](https://mlsanity.com/paper/2009.13081/) (2020)
- [PubMedQA: A Dataset for Biomedical Research Question Answering](https://mlsanity.com/paper/1909.06146/) (2019)
- [MedMCQA : A Large-scale Multi-Subject Multi-Choice Dataset for Medical domain Question Answering](https://mlsanity.com/paper/2203.14371/) (2022)
- [Capabilities of Gemini Models in Medicine](https://mlsanity.com/paper/2404.18416/) (2024)
- [AgentClinic: a multimodal agent benchmark to evaluate AI in simulated clinical environments](https://mlsanity.com/paper/2405.07960/) (2024)
- [MediQ: Question-Asking LLMs and a Benchmark for Reliable Interactive Clinical Reasoning](https://mlsanity.com/paper/2406.00922/) (2024)
- [MedFuzz: Exploring the Robustness of Large Language Models in Medical Question Answering](https://mlsanity.com/paper/2406.06573/) (2024)
- [Simulated patient systems powered by large language model-based AI agents offer potential for transforming medical education](https://mlsanity.com/paper/2409.18924/) (2024)
- [MedDialogRubrics: A Comprehensive Benchmark and Evaluation Framework for Multi-turn Medical Consultations in Large Language Models](https://mlsanity.com/paper/2601.03023/) (2026)
- [The Dialogue That Heals: A Comprehensive Evaluation of Doctor Agents' Inquiry Capability](https://mlsanity.com/paper/2509.24958/) (2025)
- [Clinical knowledge in LLMs does not translate to human interactions](https://mlsanity.com/paper/2504.18919/) (2025)
- [SycoEval-EM: Sycophancy Evaluation of Large Language Models in Simulated Clinical Encounters for Emergency Care](https://mlsanity.com/paper/2601.16529/) (2026)
- [LingxiDiagBench: A Multi-Agent Framework for Benchmarking LLMs in Chinese Psychiatric Consultation and Diagnosis](https://mlsanity.com/paper/2602.09379/) (2026)
- [Beyond Idealized Patients: Evaluating LLMs under Challenging Patient Behaviors in Medical Consultations](https://mlsanity.com/paper/2603.29373/) (2026)
- [MedPI: Evaluating AI Systems in Medical Patient-facing Interactions](https://mlsanity.com/paper/2601.04195/) (2025)

## Built on by (most cited first, 4 of 4)

- [ClinEnv: An Interactive Multi-Stage Long Horizon EHR Environment for Agents](https://mlsanity.com/paper/2606.02568/) (2026)
- [MedBench v5: A Dynamic, Process-Oriented, and Hallucination-Aware Benchmark for Clinical Multimodal Models](https://mlsanity.com/paper/2606.24155/) (2026)
- [MedRealMM: A Real-World Multimodal Benchmark for Chinese Online Medical Consultation](https://mlsanity.com/paper/2607.09142/) (2026)
- [Measuring Stability and Failure Behavior in Language Models Under Structured Perturbations](https://mlsanity.com/paper/2608.22138/) (2026)

## Abstract (arXiv)

Interactive medical dialogue benchmarks have shown that LLM diagnostic accuracy degrades significantly when interacting with non-cooperative patients, yet existing approaches either apply adversarial behaviors without graded severity or case-specific grounding, or reduce patient non-cooperation to a single ungraded axis, and none analyze cross-dimension interactions. We introduce MedDialBench, a benchmark enabling controlled, dose-response characterization of how individual patient behavior dimensions affect LLM diagnostic robustness. It decomposes patient behavior into five dimensions -- Logic Consistency, Health Cognition, Expression Style, Disclosure, and Attitude -- each with graded severity levels and case-specific behavioral scripts. This controlled factorial design enables graded sensitivity analysis, dose-response profiling, and cross-dimension interaction detection. Evaluating five frontier LLMs across 7,225 dialogues (85 cases x 17 configurations x 5 models), we find a fundamental asymmetry: information pollution (fabricating symptoms) produces 1.7-3.4x larger accuracy drops than information deficit (withholding information), and fabricating is the only configuration achieving statistical significance across all five models (McNemar p < 0.05). Among six dimension combinations, fabricating is the sole driver of super-additive interaction: all three fabricating-involving pairs produce O/E ratios of 0.70-0.81 (35-44% of eligible cases fail under the combination despite succeeding under each dimension alone), while all non-fabricating pairs show purely additive effects (O/E ~ 1.0). Inquiry strategy moderates deficit but not pollution: exhaustive questioning recovers withheld information, but cannot compensate for fabricated inputs. Models exhibit distinct vulnerability profiles, with worst-case drops ranging from 38.8 to 54.1 percentage points.