Reproduce “Video World Models Need Predictable Latents” 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
Mattia Scardecchia, Sherry Yang
The authors develop an action-conditioned video world model that combines temporal objectives for an autoencoder with a decoder conditioned on a past frame. The temporal objectives shape latents for prediction, while the decoder reuses persistent visual details. On out-of-distribution robot-video rollouts, latent prediction achieves LPIPS 0.257 versus 0.302 for reconstruction-only training; evidence is limited to synthetic data and robot-manipulation video.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.