Search papers, labs, and topics across Lattice.
1
0
2
Discarding irrelevant reasoning tokens can make language models three times faster at test time without sacrificing performance.
