Search papers, labs, and topics across Lattice.
Department of Statistics and Data Science, National Key R&D Program of China 2024YFA1015800, High Performance Computing Center, Tsinghua University. P, Tsinghua University Dushi Program, Tsinghua University
Tsinghua AI1
0
2
Pairwise evaluations of LLMs can yield cyclic inconsistencies, but a novel prompt perturbation framework can stabilize rankings and enhance interpretability.
