Search papers, labs, and topics across Lattice.
1
0
3
Chain-of-thought reasoning can be sped up by 25-50% without sacrificing accuracy, simply by watching how the model's confidence changes as it thinks.
