Reproduce “Continually Evolving Skill Knowledge in Vision Language Action Model” 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

Yuxuan Wu, Guangming Wang, Zhiheng Yang, Tianchen Deng, Maoqing Yao, Brian Sheil, Hesheng Wang

Stellar VLA addresses continual robot imitation learning by representing evolving task knowledge and routing actions through existing experts rather than adding task-specific parameters. T-Stellar models tasks, while TS-Stellar adds shared skill structure. On scratch LIBERO-long with 1% replay, TS-Stellar reached 35.0 ± 3.2% final success versus 18.9 ± 1.1% for Experience Replay; this is a simulation result.

Summary by alphaXiv.

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