Reproduce “Auriga Technical Report: A Task-Agnostic Agent Harness for Zero-Shot Embodied 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
Chengcheng Wang, Yuheng Shi, Siyu Xu, Yutong Wang, Jinjing Zhao, Chenghao Xia, Jinqi Liao, Jiankun Peng and 13 more
Auriga provides a harness for robot manipulation that previews planned motions and checks execution against declared tolerances, returning measured errors for recovery without resetting. On RoboDojo, GPT-6 Astra reaches 57.74% success rate and 64.00 Score across 6,300 episodes, narrowly ahead of the next-highest entry. A diagnostic finds the harness recovers 68.3% of the success-rate gap; results are simulated.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.