Search papers, labs, and topics across Lattice.
Affiliation:
1
0
3
Token-level LLM routing, previously hindered by system inefficiencies, now achieves up to 64x higher throughput with a novel request-centric, model-centric serving architecture.
