Reproduce “ResCue: Residual Spatial Cueing for Language-Conditioned Imitation Learning” 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
Yanzhe Tang, Xinyu Shao, Yuxuan Hu, Siyu Chen, Chenxi Zhai, Xi Cheng, Bowen Yang, Yajun Gao and 3 more
ResCue guides robot manipulation by converting a language instruction into a target mask, encoding it separately from RGB, and adding its features to spatial image features before later visual processing. The policy also uses proprioception and observation history. Across 15 simulated tasks, it averages 56.5% success, versus 40.7% for π₀; evidence is limited to tested settings and DP.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.