Reproduce “RoboCasa-Shift: Stage-Aware Robustness Evaluation of Vision-Language-Action Policies in Long-Horizon Household 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
Haocheng Wang, Youquan Liu, Weiyu Guo, Aokai Zhang, Donghang Wu, Hongyu Liu, Sha Zhang, Weidong Yang and 1 more
RoboCasa-Shift evaluates vision-language-action policy robustness in long-horizon household tasks in simulation. The authors compare clean and shifted versions of the same episodes, tracking official task success and ordered simulator-state milestones. Camera orientation L1 produced zero aggregate success-rate degradation, although eight of 80 matched pairs changed outcome. Tests covered one policy and four clean-qualified tasks, without real-world validation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.