RoboCasa-Shift: Stage-Aware Robustness Evaluation of Vision-Language-Action Policies in Long-Horizon Household Manipulation
VLA robustness benchmark on RoboCasa365: paired clean/perturbed rollouts differing in one shift, plus stage-level diagnostics of where long-horizon failures start; evaluated on GR00T N1.5.基于 RoboCasa365 的 VLA 鲁棒性诊断基准：成对比较只差一个分布偏移因素的干净/扰动执行，并定位长时序任务中失败从哪个阶段开始，而不只看总成功率；在 GR00T N1.5 上评测。
BibTeX
@misc{wang2026robocasa,
title = {{RoboCasa-Shift: Stage-Aware Robustness Evaluation of Vision-Language-Action Policies in Long-Horizon Household Manipulation}},
author = {Haocheng Wang and Youquan Liu and Weiyu Guo and Aokai Zhang and Donghang Wu and Hongyu Liu and Sha Zhang and Weidong Yang and Ben Fei},
year = {2026},
eprint = {2610.27861},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.27861},
}Datasets & Benchmarks数据集与基准VLA & Robot Foundation ModelsVLA 与机器人基础模型Sim-to-Real & Simulators仿真与 Sim-to-Realbenchmarkrobustnesslong-horizonhouseholdsim