AI · Studies · 2026-09-02 03:07:00
An arXiv paper finds that LLMs evaluating AI clinical notes exhibit a "presence verification" bias—they are good at confirming content exists but struggle to identify missing critical information—and proposes recovery methods.
Auto-generated summary for reference only. RSS