Search papers, labs, and topics across Lattice.
1
0
3
19
You can get surprisingly reliable multi-turn LLM agents by having a small, open-source critic supervise a powerful, fixed LLM actor *during* the conversation.
