Search papers, labs, and topics across Lattice.
1
0
3
Latency in voice assistants plummets by 20% by speculatively executing tool calls *during* speech recognition, not after.
