Reproduce “World Models Dream of Success: Diagnosing and Repairing Failure Insensitivity in Robot World 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

Jiuyi Xu, Xiao Hu, Meida Chen, Peng Gao, Yang Ye, Yangming Shi

CureWM repairs released robot world models by fine-tuning them on execution-verified action variants alongside their original data. It modifies actions from successful demonstrations, executes the variants, and retains both failures and successes. On LIBERO-Goal, value optimism falls from 79.55% to 30.17%, while AUROC rises from 0.495 to 0.743; false positives also rise, and improved robot control is unestablished.

Summary by alphaXiv.

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