Search papers, labs, and topics across Lattice.
1
0
3
Forget complex scheduling algorithms: multiplying KV-cache availability with load balance is surprisingly effective for LLM request routing, slashing time-to-first-token by up to 92%.
