Reproduce “FASA: Feedback-Aware Sampling Adaptation for Efficient Diffusion-Based VLA Models” 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
Yuchen Han, Jianhan Wu, Xiaoyang Qu, Lingwei Kong, Shiyi Li, Jianzong Wang
FASA adjusts a diffusion policy’s denoising budget at runtime using visual and robot-state feedback. Its range adaptor expands or contracts allowed sampling steps, while a motion measure selects a step within that range. On three ManiSkill3 tasks, mean success rose from 70.28% to 76.39%, with a 1.45× average action-generation speedup over Diffusion Policy. These are simulation results, not real-robot validation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.