Preprint
Aug 2026
AOSpec: Action and Observation Co-Speculation for Low-Latency Agent Serving
This work presents AOSpec, a lossless framework that co-speculates actions and observations across the full agent-environment loop, and outperforms every practical baseline, reducing mean end-to-end latency by 11.8-32.5% and p99 latency by up to 42.8%.
H. Chen, Jin-Nan Guo, Wayne Luk et al.
· 0 citations