Reproduce “A$^2$-VLA: Adapting Vision-Language-Action Policies to New Tools through Agentic Affordance Discovery” 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
Chenyuan Zhou, Chaoyi Huang, Zibo Wang, Zhiyuan Li, Yong-Lu Li
The authors introduce A²-VLA, which identifies task-relevant tool features, tests and revises them in reconstructed simulations, then guides a vision-language-action policy. Using teleoperated demonstrations with source tools, it reaches 86.1% mean simulated retargeting success and 35.3% mean real-robot success on unseen tools, versus 4.7% without affordance guidance. Its scope is retargetable tasks, and policy execution can still fail.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.