Jiahui Niu
Jiahui Niu is a researcher at the State Key Laboratory of Processors, Institute of Computing Technology, Chinese Academy of Sciences (ICT CAS / UCAS). Niu is the lead author of Realtime-VLA FLASH (arXiv:2605.13778), a software-level speculative inference framework for diffusion-based Vision-Language-Action models that achieves approximately 3x reduction in inference latency without retraining the base model. The central architectural insight of FLASH connects flow-matching interpolation paths to a flow-consistency-based parallel verification mechanism, replacing sequential denoising steps with cheap consistency checks against Gaussian-interpolated intermediate states.
“FLASH is a software-level framework that makes the same model run 3x faster at inference time, with almost no degradation in task success — no retraining the main model, no new hardware.”
Source→“Jiahui Niu — Institute of Computing Technology, CAS / UCAS; Project lead on FLASH. Leads the speculative inference formulation for continuous action space.”
Source→AI-extracted from podcast / newsletter / paper summaries. May contain errors.