Search papers, labs, and topics across Lattice.
UC Berkeley
1
0
3
Transformers can provably internalize chain-of-thought reasoning, matching the sample efficiency of explicit CoT while eliminating its inference overhead.