Reproduce “Task-Critical Part Alignment Improves Imitation of Object Interactions” 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

Denis Hadjivelichkov, Sebastian Ægidius, Valerio Modugno, Dimitrios Kanoulas

The method improves one-shot imitation across object instances by aligning demonstrations around the region where the tool contacted the target. It matches visual features there and uses RANSAC to reduce inconsistent correspondences before replaying the trajectory. Across seven real-world tasks, it reports a 13.8% success improvement over DINOBot, with failures remaining on reflective or low-contrast surfaces.

Summary by alphaXiv.

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