Reproduce “RARA: Robust Policy Fine-Tuning through Retrieval-Anchored Representation Alignment for Robotic 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
Xuhui Kang, Sung-Wook Lee, Yan Zhuang, Tianle Zhong, Yen-Ling Kuo
RARA adapts robot policies to manipulation tasks from limited demonstrations. It retrieves pretraining examples with similar task intent, matches motion stages, then fine-tunes on target demonstrations while anchoring retrieved-example representations. Across 13 tasks, it improves average success over target-only fine-tuning by 9.5 percentage points in-distribution and 13.6 out-of-distribution, for visual shifts with the same robot and action space.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.