Search papers, labs, and topics across Lattice.
Chinese Academy of Sciences
1
0
2
Early detection of failure in LLM agents can save over 47% of inference compute by leveraging internal representations rather than observable behavior.
