Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
MedDialBench is introduced as a benchmark to evaluate LLM diagnostic robustness by decomposing patient behavior into five dimensions (Logic Consistency, Health Cognition, Expression Style, Disclosure, and Attitude) with graded severity levels and case-specific behavioral scripts. The benchmark reveals a fundamental asymmetry where information pollution (fabricating symptoms) significantly degrades accuracy compared to information deficit (withholding information). Furthermore, fabricating symptoms is identified as the primary driver of super-additive interaction effects among different patient behavior dimensions.
LLMs are far more vulnerable to patients who lie than patients who simply withhold information, and no amount of questioning can fix it.
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.