When Correct Beliefs Collapse: Epistemic Resilience of LLMs under Clinical Pressure

A new study reveals that large language models (LLMs) frequently abandon correct medical diagnoses when subjected to escalating pressure during multi-turn clinical dialogues, despite high benchmark accuracy. Researchers introduced Med-Stress, a stress test framework, finding a significant gap between medical knowledge and belief stability across nine frontier models. To address this, the team developed RBED and R-FT, with the latter nearly eliminating belief change by training models to resist pressure.

arXiv:2605.23932v1 Announce Type: new Abstract: Despite strong medical benchmark accuracy, LLMs can exhibit severe multi-turn sycophancy in clinical dialogue, abandoning initial correct diagnosis under escalating pressure. We propose \textbf{\textsc{Med-Stress}}, a targeted stress test framework that evaluates belief stability under escalating pressure. Across nine frontier large language models LLMs , we find a clear dissociation between medical knowledge and robustness: high initial diagnostic capability does not imply high belief stability, yielding large knowledge-robustness gaps for several LLMs. To mitigate this failure mode, we propose a lightweight inference-time defense, \textbf{\texttt{RBED}} \textbf{R}ole-\textbf{B}ased \textbf{E}pistemic \textbf{D}efense , and \textbf{\texttt{R-FT}} \textbf{R}esilience-oriented \textbf{F}ine-\textbf{T}uning , a training-time approach that internalizes evidence-based resistance to pressure. Experiments show that \textbf{\texttt{R-FT}} nearly eliminates belief change and substantially improves robustness.