Search papers, labs, and topics across Lattice.
1
0
2
A structured prompting method can halve unsafe overconfidence in clinical LLMs, but the safety gains come at a steep cost to helpfulness that varies significantly by model.