Reproduce “DRAM: Delta-rule Recurrent Associative Memory for Robot Manipulation Policies” 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
Xinyu Zhao, Yixiang Shan, Tao Yang, Runyu Lei, Yiming Zhao, Jiaxin Fan, Zongbao Feng, Peng Jia
DRAM lets manipulation policies use past observations without retaining a growing image stack. It reads the current frame’s visual patches from a fixed-size associative memory, then jointly updates that memory while the pretrained vision-language backbone remains frozen. In matched LIBERO tests, DiT’s LIBERO-10 success rate rose from 75% to 91%; π0.5 tied on Goal, limiting the finding to tested settings.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.