Retrieval-augmented generation is a practical paradigm for question answering over long documents, but it remains brittle for multimodal reading where text, tables, and figures are interleaved across many pages. First, flat chunking breaks document-native structure and cross-modal alignment, yielding semantic fragments that are hard to interpret in isolation. Second, even iterative retrieval can fail in long contexts by looping on partial evidence or drifting into irrelevant sections as noise accumulates, since each step is guided only by the current snippet without a persistent global search state. We introduce $G^2$-Reader, a dual-graph system, to address both issues. It evolves a Content Graph to preserve document-native structure and cross-modal semantics, and maintains a Planning Graph, an agentic directed acyclic graph of sub-questions, to track intermediate findings and guide stepwise navigation for evidence completion. On VisDoMBench across five multimodal domains, $G^2$-Reader with Qwen3-VL-32B-Instruct reaches 66.21\% average accuracy, outperforming strong baselines and a standalone GPT-5 (53.08\%).

翻译：检索增强生成是面向长文档问答的实用范式，但在文本、表格与图表跨多页面交错排列的多模态阅读场景中仍显脆弱。首先，扁平化的分块策略会破坏文档固有结构与跨模态对齐关系，产生难以独立解读的语义片段。其次，在长上下文场景中，即使迭代检索也可能因陷入局部证据循环或随噪声累积漂移至无关段落而失效，这是因为每一步检索仅受当前片段引导而缺乏持续性的全局搜索状态。为此，我们提出G^2-Reader——一种双图系统以应对上述问题。该系统通过演化内容图来保持文档固有结构与跨模态语义，同时维护规划图（一种基于子问题的智能有向无环图）以追踪中间发现并引导逐步导航完成证据收集。在涵盖五个多模态领域的VisDoMBench评测中，搭载Qwen3-VL-32B-Instruct的G^2-Reader取得了66.21%的平均准确率，显著优于强基线模型及独立运行的GPT-5（53.08%）。