Reproduce “Youtu-Parsing: Perception, Structuring and Recognition via High-Parallelism Decoding” 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
Haoyu Cao, Kun Yin, Yunfei Wu, Bing Liu, Zhongpeng Cai, Xiaotian Li, Huang Chen, Xin Li and 11 more
Youtu-Parsing parses document pages by encoding each page into shared visual features, locating and classifying elements, and querying regions for content. It accelerates generation through parallel candidate-token prediction with verification and batching region queries. On OmniDocBench v1.5, it scores 93.22 overall versus 92.56 for PaddleOCR-VL, while table latency falls 26.82×, under reported scenario-specific conditions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.