Search papers, labs, and topics across Lattice.
AI for Science Institute
2
0
5
Confidence estimates from LLMs can be misleading when evaluating many candidates, but a new framework ensures high-probability agreement with human judgments.