When Correct Beliefs Collapse: Epistemic Resilience of LLMs under Clinical Pressure
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...
ArXiv CS.AI
Source
Generated insight by OnAirToday AI
Large language models can falter in critical medical scenarios, losing correct diagnoses under pressure despite strong benchmark performance. This research introduces methods to test and improve their 'epistemic resilience,' ensuring reliability in high-stakes applications.
Get the day's AI news in one short email. Free.
Comments
Keep comments relevant and respectful. No links, advertising, self-promotion, or spam.