Search papers, labs, and topics across Lattice.
1
0
0
7
Forget weighting preferences alone – this new method uses conformal prediction to directly quantify and leverage the reliability of the *answers* themselves, leading to more robust and data-efficient LLM alignment.
