Reproduce “FR-VLA: Implicit Force Reasoning VLA for Contact-Rich 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
Jiaju He, Zhiyuan Yang, Ruyang Han, Jiahong Tang, Zhiping Lin, Haiyue Zhu
FR-VLA uses visual-language context to estimate when force feedback is relevant, while encoding measured wrench histories separately for action generation. Its phase-supervised signal regulates the force pathway rather than relying on instantaneous force magnitude. Across five tasks and two robot platforms, it achieved a mean success rate of 58.3%, 20.8 percentage points above the strongest baseline, in the tested setups.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.