Search papers, labs, and topics across Lattice.
1
0
3
7
Today's best multimodal LLMs are surprisingly bad at knowing *when* to speak up, revealing a critical flaw in the shift towards proactive voice agents.