Search papers, labs, and topics across Lattice.
2
0
3
Even the top-performing LLM struggles with realistic user interactions, achieving only 61% success in complex task scenarios.
LLMs talk too much: HANDRAISER teaches agents to interrupt each other, slashing communication costs by 32% without sacrificing task performance.
