Reproduce “Elastic-WAM: Adaptive Spatiotemporal Imagination for Efficient Robot Control” 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
Haitian Li, Tianfu Li, Xuxi Zhao, Zizheng Liu, Yangcheng Yu, Yiwen Hou, Zeyu Xiao, Mutian Xu and 3 more
Elastic-WAM uses a single state-dependent controller to decide when to stop refining a predicted future, how far to advance denoising, and which regions to refresh. On RoboCasa-GR1, it reaches 57.4% average success with 388.2 ms latency, versus Joint-WAM at 55.0% and 572.6 ms. In tea-task trials, full completion is 50% versus 30% nominally, but both reach 30% under object replacement.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.