Search papers, labs, and topics across Lattice.
1
0
3
4
LLMs can now predict other LLMs' performance with 14% higher accuracy, even when only seeing one or two data points, by blending statistical priors with reasoning.
