Reproduce “Latent Diffusion Planning for Imitation Learning” 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

Amber Xie, Oleh Rybkin, Dorsa Sadigh, Chelsea Finn

Stanford and UC Berkeley researchers develop Latent Diffusion Planning (LDP), a framework that enables robots to learn control policies from both suboptimal trajectories and action-free demonstrations through latent space planning, achieving competitive performance on manipulation tasks while reducing dependence on expensive expert demonstrations.

Key insights

  • Dense forecasting in latent space outperforms hierarchical planning approaches
  • Combining suboptimal and action-free data leads to improved performance
  • Learned latent representations are more effective than fixed pretrained embeddings

Summary by alphaXiv.

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