DocSeeker：面向长文档理解的结构化视觉推理与证据定位
DocSeeker: Structured Visual Reasoning with Evidence Grounding for Long Document Understanding
针对长文档理解中低信噪比和监督稀缺问题，提出结构化视觉推理流程与证据感知的强化学习优化方法，可复用于需要定位关键视觉证据的多模态任务。
原文依据 we employ an Evidence-aware Group Relative Policy Optimization which jointly optimizes for both evidence localization and answer accuracy
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论现有多模态大语言模型在长文档理解任务上随着文档长度增加而出现显著性能下降。这源于两个根本挑战：1）低信噪比，关键证据淹没在无关页面中；2）监督稀缺，仅提供最终简短答案的数据集带来弱学习信号。本文通过提出一种要求模型执行结构化“分析、定位与推理”工作流的范式来应对这些挑战。为灌输这一能力，我们设计了一个两阶段训练框架：首先对通过高效知识蒸馏策略生成的高质量数据进行监督微调；随后采用证据感知的组相对策略优化，联合优化证据定位和答案准确性。此外，我们引入证据引导的分辨率分配策略，以缓解多页文档训练时的内存限制。大量实验表明，DocSeeker在域内和域外任务上均取得优越性能。我们证明其能从短页训练稳健泛化到超长文档，并与视觉检索增强生成系统自然协同，作为其实现的理想基础。
英文原文摘要 1400 字符
Existing Multimodal Large Language Models (MLLMs) suffer from significant performance degradation on the long document understanding task as document length increases. This stems from two fundamental challenges: 1) a low Signal-to-Noise Ratio (SNR), with crucial evidence buried in irrelevant pages; and 2) supervision scarcity, as datasets offering only final short answers provide a weak learning signal. In this paper, we address these challenges by proposing a paradigm that requires the model to execute a structured "Analysis, Localization and Reasoning" workflow. To instill this capability, we design a two-stage training framework: we first perform Supervised Fine-Tuning on high-quality data generated via an efficient knowledge distillation strategy. Subsequently, we employ an Evidence-aware Group Relative Policy Optimization which jointly optimizes for both evidence localization and answer accuracy. Additionally, we introduce a Evidence-Guided Resolution Allocation strategy to mitigate memory constraints of training on multi-pages documents. Extensive experiments demonstrate that DocSeeker achieves superior performance on both in-domain and out-of-domain tasks. We show it robustly generalizes from short-page training to ultra-long documents and is naturally synergistic with visual Retrieval-Augmented Generation systems, serving as an ideal foundation for their implementation.