Realtime-VLA FLASH: Speculative Inference Framework for Diffusion-based VLAs
Jiahui Niu, Kefan Gu, Yucheng Zhao, Shengwen Liang, Tiancai Wang + 3 more
TLDR
Realtime-VLA FLASH uses speculative inference with a lightweight draft model to significantly reduce latency in diffusion-based VLAs for real-time embodied tasks.
Key contributions
- Introduces Realtime-VLA FLASH, a speculative inference framework for dVLAs.
- Uses a lightweight draft model with parallel verification by the main model's Action Expert.
- Features a phase-aware fallback mechanism for reliable low-latency replanning.
- Achieves 3.04x speedup on LIBERO, reducing average inference latency to 19.1 ms.
Why it matters
Diffusion-based VLAs are powerful but too slow for real-time robotics. This paper addresses a fundamental limitation by enabling high-frequency, low-latency replanning. It opens doors for deploying advanced embodied intelligence in practical, latency-critical applications.
Original Abstract
Diffusion-based vision-language-action models (dVLAs) are promising for embodied intelligence but are fundamentally limited in real-time deployment by the high latency of full inference. We propose Realtime-VLA FLASH, a speculative inference framework that eliminates most full inference calls during replanning by introducing a lightweight draft model with parallel verification via the main model's Action Expert and a phase-aware fallback mechanism that reverts to the full inference pipeline when needed. This design enables low-latency, high-frequency replanning without sacrificing reliability. Experiments show that on LIBERO, FLASH largely preserves task performance by replacing many 58.0 ms full-inference rounds with speculative rounds as fast as 7.8 ms, lowering task-level average inference latency to 19.1 ms (3.04x speedup). We additionally demonstrate effectiveness on real-world conveyor-belt sorting, highlighting its practical impact for latency-critical embodied tasks.
📬 Weekly AI Paper Digest
Get the top 10 AI/ML arXiv papers from the week — summarized, scored, and delivered to your inbox every Monday.