Reproduce “SURE:Spatiotemporal Uncertainty Representation for Efficient Vision-Language-Action Learning” 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
Jingye Zhang, Choo Sin Wai, Dake Zhong, Zhifang Liu, Yuhong Zhang, Siyao Xiao, Guorui Song, Guocun Wang and 3 more
SURE addresses pointwise robot imitation’s penalty for valid differences in pose and timing. Its probabilistic action decoder models neighboring action bins as related possibilities, while Soft-DTW aligns action sequences across local timing shifts. On LIBERO-Spatial, SURE achieves 95.0% success with 25% of the demonstrations, versus 94.4% for OpenVLA-OFT with all data, at 10,000 training steps. The comparison is single-seed.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.