Reproduce “Trajectory-Calibrated Merging of Embodied 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

Wenshuo Chen, Honglei Jin, Le Yang, Jianfei Song, Shaofeng Liang, Lei Wang, Junze Zhu, Kailong Wang and 6 more

The authors introduce Trajectory-Calibrated Regression (TCR), which merges four robot specialists into one fixed policy. It calibrates the merged model using expert execution trajectories, intermediate action-generation calls, and features recomputed through the already-merged network. On LIBERO, TCR achieves 77.83% task-macro success versus 68.33% for FeatCal, leading among static mergers tested on this four-expert setup and protocol.

Summary by alphaXiv.

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