Reproduce “FailureGen: Scalable Robot Failure Data Generation with 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

Akul Santhosh, Chaitra Sathe, Aastha Jhunjhunwala

FailureGen creates failure video examples by transferring motion from recorded failures into successful scenes and generating continuations with an adapted world model. It labels clips by observed outcomes and filters examples. In an evaluation, video-only detection identified 33 of 36 failure intervals, versus 28 of 36 with real-only training; differing training configurations mean the comparison does not isolate augmentation.

Summary by alphaXiv.

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