Search papers, labs, and topics across Lattice.
2
0
5
Specialization in vision-language models dramatically increases the risk of re-linking de-identified medical images to their original reports, revealing critical privacy vulnerabilities in clinical AI applications.
Scaling clinical LLMs doesn't guarantee safety: high-risk errors persist even with advanced RAG and max-context prompting, highlighting the critical role of evidence quality and deployment strategy.
