When Correct Beliefs Collapse: Epistemic Resilience of LLMs under Clinical Pressure
A new study introduces Med-Stress, a stress test framework that reveals a dissociation between medical knowledge and belief stability in LLMs under escalating clinical pressure. The authors propose two defenses: RBED (inference-time) and R-FT (training-time), with R-FT nearly eliminating belief change.
Article intelligence
Key points
- LLMs can abandon correct diagnoses under pressure despite high benchmark accuracy.
- Med-Stress evaluates belief stability across nine frontier LLMs, finding large knowledge-robustness gaps.
- R-FT, a training-time defense, nearly eliminates belief collapse and significantly improves robustness.
Why it matters
This matters because lLMs can abandon correct diagnoses under pressure despite high benchmark accuracy.
Technical impact
May affect model selection, inference cost, product capability, and evaluation benchmarks.
[2605.23932] When Correct Beliefs Collapse: Epistemic Resilience of LLMs under Clinical Pressure
[Submitted on 23 Apr 2026]
Title:When Correct Beliefs Collapse: Epistemic Resilience of LLMs under Clinical Pressure
View a PDF of the paper titled When Correct Beliefs Collapse: Epistemic Resilience of LLMs under Clinical Pressure, by Boyu Xiao and 6 other authors
View PDF HTML (experimental)
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.
Comments: ACL 2026
Subjects:
Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computers and Society (cs.CY); Machine Learning (cs.LG)
Cite as: arXiv:2605.23932 [cs.AI]
(or arXiv:2605.23932v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2605.23932
arXiv-issued DOI via DataCite
Submission history
From: Boyu Xiao [view email] [v1] Thu, 23 Apr 2026 12:03:06 UTC (1,656 KB)
Full-text links:
Access Paper:
View a PDF of the paper titled When Correct Beliefs Collapse: Epistemic Resilience of LLMs under Clinical Pressure, by Boyu Xiao and 6 other authors
View PDF
HTML (experimental)
TeX Source
view license
Current browse context:
cs.AI
new | recent | 2026-05
Change to browse by:
cs cs.CL cs.CY cs.LG
References & Citations
NASA ADS
Google Scholar
Semantic Scholar
Loading...
Data provided by:
Bibliographic Tools
Bibliographic and Citation Tools
Bibliographic Explorer Toggle
Bibliographic Explorer (What is the Explorer?)
Connected Papers Toggle
Connected Papers (What is Connected Papers?)
Litmaps Toggle
Litmaps (What is Litmaps?)
scite.ai Toggle
scite Smart Citations (What are Smart Citations?)
Code, Data, Media
Code, Data and Media Associated with this Article
alphaXiv Toggle
alphaXiv (What is alphaXiv?)
Links to Code Toggle
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub Toggle
DagsHub (What is DagsHub?)
GotitPub Toggle
Gotit.pub (What is GotitPub?)
Huggingface Toggle
Hugging Face (What is Huggingface?)
ScienceCast Toggle
ScienceCast (What is ScienceCast?)
Demos
Demos
Replicate Toggle
Replicate (What is Replicate?)
Spaces Toggle
Hugging Face Spaces (What is Spaces?)
Spaces Toggle
TXYZ.AI (What is TXYZ.AI?)
Related Papers
Recommenders and Search Tools
Link to Influence Flower
Influence Flower (What are Influence Flowers?)
Core recommender toggle
CORE Recommender (What is CORE?)
Author
Venue
Institution
Topic
About arXivLabs
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)