Reproduce “Target-Consistent Action Guidance via Semi-Unbalanced Optimal Transport for Drifting World Models” 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

Md Abtahi Majeed Chowdhury, Md Rifat Ur Rahman, Md Sakib Hossain Shovon, Jaesik Choi, Minhyuk Sung

MSA changes action guidance in one-step world models by subtracting the guidance a correct recorded future would receive, then applying a shared translation. In a controlled T-block toy at scale 2.5, it reaches progress 1.001 with endpoint error .003, while balanced W-Flow reaches 2.504. On Push-T, Kendall τ is .712 versus .693; benchmark models under-react.

Summary by alphaXiv.

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