arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2604.04771cs.CVcs.CL
收藏

MinerU2.5-Pro: 推动大规模数据导向文档解析的极限

MinerU2.5-Pro: Pushing the Limits of Data-Centric Document Parsing at Scale

Shanghai Artificial Intelligence Laboratory(上海人工智能实验室);Peking University(北京大学);Shanghai Jiao Tong University(上海交通大学)
Bin Wang, Tianyao He, Linke Ouyang, Fan Wu, Zhiyuan Zhao, Tao Chu, Yuan Qu, Zhenjiang Jin, Weijun Zeng, Ziyang Miao, Bangrui Xu, Junbo Niu, Mengzhang Cai, Jiant… 展开作者

Bin Wang, Tianyao He, Linke Ouyang, Fan Wu, Zhiyuan Zhao, Tao Chu, Yuan Qu, Zhenjiang Jin, Weijun Zeng, Ziyang Miao, Bangrui Xu, Junbo Niu, Mengzhang Cai, Jiantao Qiu, Qintong Zhang, Dongsheng Ma, Yuefeng Sun, Hejun Dong, Wenzheng Zhang, Jutao Xiao, Jiayong Shi, Pengyu Liao, Xiaomeng Zhao, Huaping Zhong, Liqun Wei, Jing Yu, Jie Yang, Wei Li, Shasha Wang, Qianqian Wu, Xuanhe Zhou, Weijia Li, Zhenxiang Li, Zhongying Tu, Jiang Wu, Lijun Wu, Chao Xu, Kai Chen, Wentao Zhang, Yu Qiao, Bowen Zhou, Dahua Lin, Conghui He

AI总结:

本文提出MinerU2.5-Pro,通过数据工程和训练策略设计提升文档解析性能,无需改变模型架构,在OmniDocBench v1.6上取得95.69的高分。

摘要:

当前文档解析方法主要通过模型架构创新进步,而系统性的训练数据工程仍被忽视。然而,跨架构和参数规模的最先进模型在相同困难样本集上表现出高度一致的失败模式,表明性能瓶颈源于训练数据的共同缺陷而非架构差异。基于此发现,我们提出MinerU2.5-Pro,通过数据工程和训练策略设计提升性能,保留MinerU2.5的1.2B参数架构不变。其核心是围绕覆盖性、信息性和标注准确性共同设计的数据引擎:多样性与难度意识采样将训练数据从不足1000万扩展到6550万样本,缓解分布偏移;跨模型一致性验证利用异构模型的输出共识评估样本难度并生成可靠标注;Judge-and-Refine流程通过渲染后验证的迭代修正提升困难样本的标注质量。三阶段渐进训练策略——大规模预训练、困难样本微调和GRPO对齐——依次利用不同质量层级的数据。在评估方面,我们修正OmniDocBench v1.5中的元素匹配偏差并引入困难子集,建立更具判别性的OmniDocBench v1.6协议。在不修改任何架构的情况下,MinerU2.5-Pro在OmniDocBench v1.6上达到95.69,优于相同架构基线2.71分,并超越所有现有方法,包括基于参数规模超200倍的模型方法。

英文摘要:

Current document parsing methods advance primarily through model architecture innovation, while systematic engineering of training data remains underexplored. Yet state-of-the-art models spanning diverse architectures and parameter scales exhibit highly consistent failure patterns on the same set of hard samples, suggesting that the performance bottleneck stems from shared deficiencies in training data rather than from architectural differences. Building on this finding, we present MinerU2.5-Pro, which advances the state of the art purely through data engineering and training strategy design while retaining the 1.2B-parameter architecture of MinerU2.5 unchanged. At its core is a Data Engine co-designed around coverage, informativeness, and annotation accuracy: Diversity-and-Difficulty-Aware Sampling expands training data from under 10M to 65.5M samples while mitigating distribution shift; Cross-Model Consistency Verification leverages output consensus among heterogeneous models to assess sample difficulty and generate reliable annotations; the Judge-and-Refine pipeline improves annotation quality for hard samples through render-then-verify iterative correction. A three-stage progressive training strategy--large-scale pre-training, hard sample fine-tuning, and GRPO alignment--sequentially exploits these data at different quality tiers. On the evaluation front, we rectify element-matching biases in OmniDocBench v1.5 and introduce a Hard subset, establishing the more discriminative OmniDocBench v1.6 protocol. Without any architectural modification, MinerU2.5-Pro achieves 95.69 on OmniDocBench v1.6, improving over the same-architecture baseline by 2.71 points and surpassing all existing methods, including those based on models with over 200x more parameters.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑