Infinity-Parser2 Technical Report

Abstract: We present Infinity-Parser2, a large multimodal model that couples a controllable data-synthesis pipeline with multi-task reinforcement learning for end-to-end document parsing, addressing the persistent scarcity of faithfully annotated parsing corpora.

摘要： 我们提出了 Infinity-Parser2，这是一个大型多模态模型。该模型将可控的数据合成流水线与多任务强化学习相结合，用于端到端的文档解析，旨在解决高质量标注解析语料库长期匮乏的问题。

Our contributions are threefold. First, we build a scalable synthesis engine, pairing a controllable rendering framework with an iterative refinement loop, and use it to construct and open-source Infinity-Doc2-5M: a 5-million-sample bilingual (Chinese/English) corpus spanning diverse document types, annotated with element bounding boxes, canonical content forms (Markdown, HTML, LaTeX, SMILES, structured charts), and full-page reading order.

我们的贡献主要体现在三个方面。首先，我们构建了一个可扩展的合成引擎，将可控渲染框架与迭代优化循环相结合，并利用它构建并开源了 Infinity-Doc2-5M：这是一个包含 500 万样本的双语（中/英）语料库，涵盖了多种文档类型，并标注了元素边界框、规范内容格式（Markdown、HTML、LaTeX、SMILES、结构化图表）以及全页阅读顺序。

Second, we introduce a verifiable, multi-task reward system that enables Joint Reinforcement Learning across eight co-trained objectives (document parsing, layout analysis, table parsing, math formula parsing, chart parsing, chemical formula parsing, document VQA, and general multimodal understanding), unifying perception, structure, and reasoning in a single optimization signal.

其次，我们引入了一个可验证的多任务奖励系统，实现了跨八个协同训练目标（文档解析、版面分析、表格解析、数学公式解析、图表解析、化学公式解析、文档视觉问答以及通用多模态理解）的联合强化学习，将感知、结构和推理统一在一个优化信号中。

Third, we release two variants under a shared architecture: Infinity-Parser2-Flash, optimized for low-latency inference with a $3.68\times$ throughput gain over Infinity-Parser-7B, and Infinity-Parser2-Pro, engineered for precision-critical settings.

第三，我们在共享架构下发布了两个变体：针对低延迟推理优化的 Infinity-Parser2-Flash（吞吐量较 Infinity-Parser-7B 提升了 3.68 倍），以及专为高精度需求场景设计的 Infinity-Parser2-Pro。

Infinity-Parser2-Pro reaches state-of-the-art 87.6% on olmOCR-Bench and 74.3% on ParseBench, surpassing DeepSeek-OCR-2, PaddleOCR-VL-1.5, and MinerU2.5, with strong generalization to charts, chemical formulas, and document VQA.

Infinity-Parser2-Pro 在 olmOCR-Bench 上达到了 87.6% 的业界领先水平，在 ParseBench 上达到了 74.3%，超越了 DeepSeek-OCR-2、PaddleOCR-VL-1.5 和 MinerU2.5，并在图表、化学公式和文档视觉问答方面表现出强大的泛化能力。