Beyond Idealized Patients: Evaluating LLMs under Challenging Patient Behaviors in Medical Consultations
This addresses the safety gap in medical LLM evaluations for real-world clinical scenarios, though it is incremental as it builds on existing datasets and focuses on a specific domain.
The paper tackles the problem of evaluating large language models (LLMs) in medical consultations under challenging patient behaviors like contradictions and inaccuracies, resulting in the creation of CPB-Bench, a benchmark of 692 dialogues, and finding that models show consistent failure patterns, especially with contradictory information, while interventions yield inconsistent improvements.
Large language models (LLMs) are increasingly used for medical consultation and health information support. In this high-stakes setting, safety depends not only on medical knowledge, but also on how models respond when patient inputs are unclear, inconsistent, or misleading. However, most existing medical LLM evaluations assume idealized and well-posed patient questions, which limits their realism. In this paper, we study challenging patient behaviors that commonly arise in real medical consultations and complicate safe clinical reasoning. We define four clinically grounded categories of such behaviors: information contradiction, factual inaccuracy, self-diagnosis, and care resistance. For each behavior, we specify concrete failure criteria that capture unsafe responses. Building on four existing medical dialogue datasets, we introduce CPB-Bench (Challenging Patient Behaviors Benchmark), a bilingual (English and Chinese) benchmark of 692 multi-turn dialogues annotated with these behaviors. We evaluate a range of open- and closed-source LLMs on their responses to challenging patient utterances. While models perform well overall, we identify consistent, behavior-specific failure patterns, with particular difficulty in handling contradictory or medically implausible patient information. We also study four intervention strategies and find that they yield inconsistent improvements and can introduce unnecessary corrections. We release the dataset and code.