Reproduce “BiCAT: Bimanual Coordination Action Tokenization” 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

Xuanwei Liu, Guipu Cao, Zhongyi Luo, Jiafei Duan

The authors introduce BiCAT, a bimanual action tokenizer that gives each arm separate tokens while incorporating partner-arm context. It also has a policy predict one arm’s token block before the other’s. In RoboTwin 2.0, BiCAT reaches 69.3% mean success, versus 51.7% for the strongest baseline. The ten-task evaluation uses tasks included in training, not unseen tasks.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.