FlashVLA: Streaming Action Decoding for Fast and Asynchronous VLA Inference
2026-08-27 • Robotics
Robotics
AI summaryⓘ
The authors address challenges in Vision-Language-Action (VLA) models used for controlling robots, specifically problems with slow action prediction and unsteady action timing. They propose FlashVLA, a new method that decodes robot actions in smaller, manageable chunks while keeping the actions smooth and continuous. This allows the robot to respond faster and more reliably, achieving over 30 updates per second on a single GPU. Their tests show FlashVLA improves speed without losing accuracy in both simulated and real-world tasks.
Vision-Language-Action modelsrobotic manipulationinference latencyasynchronous executionflow-matchingaction decodingcausal attentionautoregressive modelscontrol frequency
Authors
Zekai Li, Jiaming Tang, Zhijian Liu
Abstract
Vision-Language-Action (VLA) models are increasingly promising for robotic manipulation, yet their real-world deployment remains bottlenecked by high inference latency and unstable asynchronous execution. This challenge is particularly pronounced in flow-matching-based VLA models, where action decoding requires multiple iterative steps conditioned on the VLM context. While efficient inference methods improve control frequency and asynchronous methods reduce execution idle time, existing approaches often fail to jointly achieve low-latency inference and accurate, temporally consistent asynchronous execution. We introduce \textbf{FlashVLA}, a streaming action decoding framework that addresses both challenges in a unified formulation. FlashVLA maintains a streaming action buffer with multiple chunks at different noise levels and decodes them using chunk-wise causal attention. This design allows FlashVLA to produce one executable action chunk per inference step. Moreover, its chunk-wise autoregressive formulation implicitly preserves action continuity, enabling smooth asynchronous execution without extra future-state conditioning. Across extensive simulated and real-world experiments, FlashVLA substantially improves inference speed while maintaining strong task performance. It can achieve $\geq$30\,Hz control frequency on a single GPU with smooth asynchronous inference in real-world deployment.