Reproduce “EVA: Aligning Video World Models with Executable Robot Actions via Inverse Dynamics Rewards” 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

Ruixiang Wang, Qingming Liu, Yueci Deng, Guiliang Liu, Zhen Liu, Kui Jia

EVA uses a frozen inverse dynamics model (IDM) to decode generated robot videos into commands and provide feedback for post-training the video generator. It rewards smooth motion and penalizes velocity and acceleration limit violations, without executing every rollout. In RoboTwin 2.0’s clean simulation setting, mean success across 21 tasks rises from 45.81% to 52.29%; not proof smoothness caused the gain.

Summary by alphaXiv.

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