Reproduce “VLALagBench: Understanding When and Why VLA Inference Latency Matters” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Qi Wen, Borui Xu, Hussein Amro, Hongshi Tan, Hanmin Lin, Chaoyi Ruan, Shanghao Liu, Gang Yang and 6 more

VLALagBench measures how inference latency affects robot success on moving-target tasks, comparing synchronous, lazy asynchronous, and continuous asynchronous execution across policies and schedules. In simulation, a 100 ms delay reduced π₀.₅ set-down success from 71.5% to 40.5%; asynchronous execution sometimes recovered success but required more frequent inference requests. These results describe tested conditions, not a general guarantee for robots.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.