Search papers, labs, and topics across Lattice.
Y. Zhang, Y. Yan, N. Yang and D. Yuan are with the School of Electrical and Computer Engineering, The University of Sydney, Darlington, NSW 2008, Australia. (Email: yuning.zhang1@sydney.edu.au; yyan0244@uni.sydney.edu.au; n.yang@sydney.edu.au; dong.yuan@sydney.edu.au). The corresponding author is D. Yuan
1
0
3
0
AgentServe achieves up to 2.8x improvement in time-to-first-token and 2.7x in tokens-per-output-token for agentic workloads on a single GPU by strategically isolating prefills and decodes.
