Reproduce “Unlimited OCR Works 2” 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

Shu Wei, Jingjing Wu, Qunyi Xie, Lingshu Zhang, Xiangcheng Du, Yannian Fu, Mengjie Wang, Zuyi Wang and 5 more

Unlimited OCR 2 transcribes long documents into ordered text and structure while limiting visual attention. It combines SR-SWA, which selects fixed remote visual blocks alongside local context, with overlapping chunks that reduce visual-cache residency. On MarathonDocBench, it scores 77.06 versus 33.74 for Kimi-K3; at 100 pages, it reports 7.0× and 5.6× per-step speedups, under reported conditions.

Summary by alphaXiv.

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