Reproduce “Flide: Accelerating Decision Inference with Exact State Forking” 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
Bobo Li, Jiajia Song, Hao Fei, Wynne Hsu, Mong Li Lee
Flide accelerates decision inference by sharing a document or agent-history context across independent question branches in one flattened token stream. It preserves standalone branch inputs using attention masks and saved recurrent states, with exact equivalence in exact arithmetic. On Qwen3.5-0.8B QuALITY, latency was 37.7 ms versus 359 ms for SGLang; this is one workload result.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.