Reproduce “ProWorld: Multi-View World Model with Proprioceptive Feedback for Closed-Loop Policy Evaluation” 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
Zijun Wang, Tianyi Yan, Haokun Lin, Panwen Hu, Bohang He, Siyu Ren, Xiaoyu Guo, Weijia Liufu and 7 more
ProWorld evaluates policies in a learned world model by returning predicted camera views and proprioceptive state to a frozen policy after each action. It trains on simulated failures as well as demonstrations. Across 12 policy–task pairs in an evaluation, generated and real success rates correlate at Pearson r = 0.95 and Spearman ρ = 0.94, though long rollouts can accumulate drift.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.