用最小费用流优化多模态长文档问答中的证据整合。
FLOWREADER: Min-Cost Flow Optimization for Multi-Modal Long Document Q&A
- 将证据组装建模为多模态节点图上的最小费用流问题，统一控制选择与路由。
- 在纸面表格和幻灯片问答任务上分别提升1.30和0.62点，整体性能接近最强基线。
- 适合处理证据碎片化严重的多模态长文档，尤其适合需要跨模态推理的场景。
长篇多模态文档迫使检索增强系统从文本、表格和幻灯片中拼接答案，而这些证据常被拆分在表格单元格、多张幻灯片或图与说明之间。传统的Top-k片段独立检索无法体现证据间的关联性。我们提出FLOWREADER，将证据组装重新建模为多模态节点图上的最小费用流问题：一个单一评分向量 $h$ 控制源选择（通过MMR）、目标选择（通过长度感知的答案可回答性代理）以及每条边的成本与容量。最优流被分解为候选证据路径，通过熵正则化的复制动态选择紧凑且无冗余的子集，再由并行视觉语言模型工作者在双过程门控下生成答案，仅在答案一致性低或流路压力大时触发一次系统2精炼。在VisDoMBench上，FLOWREADER在以碎片化证据为主的PaperTab（58.40，优于G²-Reader的57.10）和SlideVQA（72.93，优于G²-Reader的72.31）两个子集上表现最佳，在SPIQA、FetaTab和SciGraphQA上也具竞争力。五个子集的宏平均得分达到65.47，仅比最强基线G²-Reader（66.21）低0.74。结果表明，最小费用流在碎片化多模态证据场景下显著优于Top-k检索，并提供了一种统一控制评分、路由、选择与自适应计算的方法。
原文摘要 · Abstract (English)
Long, multimodal documents force retrieval-augmented systems to assemble answers from evidence fragmented across text, tables, and slides broken across cells in a long table, spread over multiple slides, or split between a figure and its discussion. Top-$k$ chunk retrieval treats each fragment independently and cannot represent how evidence connects. We introduce FLOWREADER, which reframes evidence assembly as a min-cost flow problem on a multimodal node graph: a single scoring vector $h$ controls source selection (via MMR), sink selection (via a length-aware answerability proxy), and the costs and capacities of every edge. The optimal flow is decomposed into candidate evidence paths, a compact non-redundant subset is selected by entropy-regularized replicator dynamics, and parallel VLM workers under a dual-process gate produce the answer with a single System-2 refinement pass triggered when answer consistency is low or the routed flow is strained. On VisDoMBench, FLOWREADER is best on the two subsets dominated by fragmented evidence PaperTab ($58.40$, $+1.30$ over G^{2}-Reader) and SlideVQA ($72.93$, $+0.62$) and competitive on SPIQA, FetaTab, and SciGraphQA. Macro-averaged across all five subsets, FLOWREADER ($65.47$) is within $0.74$ of the strongest baseline (G^{2}-Reader, $66.21$). Overall, these results show that min-cost flow performs well on fragmented multimodal evidence, where top-$k$ retrieval fails. It also provides a unified way to control scoring, routing, selection, and adaptive compute together.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。