The paper introduces SelfDx, a benchmark that tests whether AI models appropriately respond to health queries by either cautioning against self-diagnosis or agreeing when evidence rules out serious alternatives. It evaluates 27 frontier models on 100 paired cases, finding that most models inappropriately agree in caution-required scenarios, achieving only a 77.0% appropriate caution rate. The study shows that prompt engineering can raise caution but at the cost of agreement, while supervised fine‑tuning on balanced examples improves both metrics.
arXiv:2608. 03731v1 Announce Type: new Abstract: Patient-facing medical LLMs and agents increasingly answer symptom questions before clinician contact, where the key safety question is what action the user should take next.
arXiv:2601. 17642v2 Announce Type: replace Abstract: Safety alignment in Large Language Models is critical for healthcare; however, reliance on binary refusal boundaries often results in over-refusal of benign queries or unsafe compliance with harmful ones.
arXiv:2607. 28677v1 Announce Type: new Abstract: LLM now pass medical licensing examinations and, in curated cases, can rival physicians at diagnostic reasoning.
arXiv:2608. 10258v1 Announce Type: cross Abstract: Large language models (LLMs) increasingly provide conversational health information that may influence treatment decisions, yet existing benchmarks do not isolate whether medication-safety boundaries persist across follow-ups after explicit self-treatment intent.
AcuityBench is a new benchmark that tests whether language models can correctly identify the urgency of medical care needed from user presentations. It unifies five public datasets—user conversations, online forum posts, clinical vignettes, and patient portal messages—under a shared four-level acuity framework, providing 914 cases for evaluation. The benchmark supports both explicit four-way classification and free-form conversational responses, revealing that models vary widely in accuracy and that conversational formats reduce over-triage but increase under-triage, especially for high-acuity cases.
arXiv:2508. 00923v3 Announce Type: replace Abstract: Large language models (LLMs) are increasingly used to answer health-related questions and support healthcare workflows, yet evidence for their safety still relies heavily on static benchmarks that can rapidly become obsolete or be optimized against.
arXiv:2601. 16529v4 Announce Type: replace Abstract: Large language models (LLMs) deployed in clinical decision support may acquiesce to patient requests for care that conflicts with evidence-based guidelines.
arXiv:2607. 25485v1 Announce Type: new Abstract: Health AI is evolving from answering questions to agentic systems that converse with patients, reason about health records, and act on their behalf.
arXiv:2608.21721v1 Announce Type: new Abstract: Correcting health misinformation in dialogue requires more than producing a factual rebuttal: users differ in what they know, what they believe, and wh...
arXiv:2608.29241v1 Announce Type: new Abstract: Clinical voice agents are now deployed in routine care, where real patients do not wait their turn: they interrupt. These systems typically use a casca...
The paper introduces Expected‑Severity‑Risk (ESR), a new objective for selecting questions in proactive medical dialogue that prioritizes reducing the expected severity of diagnostic errors rather than merely uncertainty. ESR uses population statistics to marginalize over possible answers and distills its rankings into a prefix‑only language policy, enabling deployment without teacher‑side risk computation. Experiments on DDxPlus show ESR cuts high‑severity diagnostic misses by 29.5% and boosts accuracy while adding only 0.14 extra questions per dialogue.
arXiv:2607. 13036v1 Announce Type: cross Abstract: Large language models (LLMs) are increasingly used for decision support in healthcare, but clinical evidence is often incomplete or evolving.