Search papers, labs, and topics across Lattice.
1
0
3
2
LLMs can slash token usage by 70% and boost reasoning accuracy by 14.8% in long-horizon tasks simply by learning when to remember and forget intermediate thoughts.
