Reproduce “SurgTransfer: Multi-Step Latent Action Modeling for Action Transfer in Surgical World Model” 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
Saurabh Koju, Love Panta, Deepak Thapa, Rudra P. K. Poudel, Prajwal Ghimire, Angus Watson, Yash Raj Shrestha, Prashnna Kumar Gyawali and 1 more
SurgTransfer transfers multi-step actions learned from unlabeled surgical video to a compatible target scene, generating a possible video continuation. Its latent action model represents sequences from an initial frame, and a video diffusion model uses those actions with the target frame to generate the result. On SurgToolLoc, it reports CD-FVD 141.41 and ECS 0.67, outperforming AdaWorld, within matched contexts.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.