Reproduce “PhysVAM: Physics-oriented Future Value Action Model for Dynamic Object Manipulation” 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

Yuxiang Yang, Jingbo Zhang, Jiaru Zhong, Junliang Chen, Yirui Zhou, Guangming Wang, Haitao Lin, Huaiyuan Xu and 2 more

PhysVAM guides dynamic manipulation by predicting task progress at multiple future offsets and using latent features to condition action generation. It predicts values 4, 8, 16, and 32 steps ahead. In evaluated settings, it reaches 58.83% simulation success versus PUMA’s 31.83%, and 64.17% real-world scenario-macro success; results are limited to the tested tasks.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.