Search papers, labs, and topics across Lattice.
1
0
2
LLMs' chain-of-thought reasoning is more reliable when the uncertainty (entropy) decreases consistently at each step, not just overall.
