Search papers, labs, and topics across Lattice.
1
0
3
LLM reasoning can be compressed without sacrificing accuracy by selectively encouraging exploration only on difficult questions, preventing premature entropy collapse.
