Reproduce “HarnessWAM: Bridging Prediction and Deliberation in World Action Models” 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

Zhaopeng Gu, Bingke Zhu, Tianxi Lin, Guibo Zhu, Yingying Chen, Kai Wang, Tingyu Yuan, Fan Yang and 4 more

HarnessWAM coordinates long-horizon robot tasks by pairing a world action model with an external task manager. The manager retains evidence, builds and projects a graph of supported skills, and verifies outcomes before continuing or revising actions. On RoboMemArena, it reports 64.4% full-task success versus 49.0% for direct LingBot-VA, using different checkpoints; results cover simulation and two physical UR3 tasks.

Summary by alphaXiv.

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