Search papers, labs, and topics across Lattice.
1
0
2
4
Models are substantially better at pairwise self-verification than independent scoring, unlocking a more efficient and accurate approach to test-time scaling for complex reasoning.
