Reproduce “AffordanceVLA: A Vision-Language-Action Model Empowering Action Generation through Affordance-Aware Understanding” 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

Qize Yu, Jiadi You, Yuran Wang, Jiaqi Liang, Bowen Ping, Yang Tian, Yue Chen, Bohan Zhou and 7 more

AffordanceVLA predicts task-specific affordances to connect robot instructions with physical actions. Its Which2Act, Where2Act, and How2Act representations identify an object, locate a 2D contact area, and encode 3D geometry, conditioning an action policy. On CALVIN ABC→D, it completes an average 4.33-task chain, versus 3.89 when affordance dimensions cannot interact; this supports joint refinement, not causal direction.

Summary by alphaXiv.

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