Search papers, labs, and topics across Lattice.
1
0
3
A self-evolving critic can reduce confidence estimation errors in LLM agents by up to 54% without requiring any training or ground truth labels.
