When Correct Beliefs Collapse: Epistemic Resilience of LLMs under Clinical Pressure
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866918519179837440 |
|---|---|
| author | Xiao, Boyu Tian, Xiuqi Song, Xuwen Wang, Haochun Song, Guanchun Zhao, Sendong Qin, Bing |
| author_facet | Xiao, Boyu Tian, Xiuqi Song, Xuwen Wang, Haochun Song, Guanchun Zhao, Sendong Qin, Bing |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_23932 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | When Correct Beliefs Collapse: Epistemic Resilience of LLMs under Clinical Pressure Xiao, Boyu Tian, Xiuqi Song, Xuwen Wang, Haochun Song, Guanchun Zhao, Sendong Qin, Bing Artificial Intelligence Computation and Language Computers and Society Machine Learning 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. |
| title | When Correct Beliefs Collapse: Epistemic Resilience of LLMs under Clinical Pressure |
| topic | Artificial Intelligence Computation and Language Computers and Society Machine Learning |
| url | https://arxiv.org/abs/2605.23932 |