用智能体框架实现文档问答的精准答案定位，兼顾准确率与可解释性。
ARIAL: An Agentic Framework for Document VQA with Precise Answer Localization
- 通过大模型规划调度多种专用工具，分步完成文本提取、上下文检索、答案生成和区域定位。
- 在四个基准上达到最高文本准确率（如DocVQA达88.7 ANLS）和空间定位精度（如FUNSD达50.3 mAP）。
- 模块化设计透明可审计，适合需要高可信度的金融、医疗等文档智能场景。
文档视觉问答（Document VQA）要求模型不仅准确提取文本答案，还需精确定位其在文档图像中的位置，这对高风险应用的可解释性至关重要。现有系统在文本准确率与空间定位可靠性之间难以兼顾。本文提出ARIAL（Agentic Reasoning for Interpretable Answer Localization），一个基于大模型规划代理的模块化框架，通过协调多种专用工具实现精准答案提取与可靠空间定位。ARIAL将文档VQA分解为：使用TrOCR进行OCR文本提取，借助语义搜索实现增强型上下文选择，采用微调的Gemma 3-27B生成答案，并通过文本到区域对齐明确标注答案位置。该架构产生清晰的推理轨迹，支持工具级可审计性和独立优化。我们在四个基准（DocVQA、FUNSD、CORD、SROIE）上评估，采用文本准确率（ANLS）和空间精度（mAP at IoU 0.50 to 0.95）。ARIAL在所有数据集上均达最优表现：DocVQA上为88.7 ANLS和50.1 mAP，FUNSD上为90.0 ANLS和50.3 mAP，CORD上为85.5 ANLS和60.2 mAP，SROIE上为93.1 ANLS，相较此前最佳方法DLaVA在DocVQA上提升+2.8 ANLS和+3.9 mAP。结果表明，专业化工具的智能体协同能同时提升性能与可解释性，为可信文档智能系统提供可行路径。
原文摘要 · Abstract (English)
Document Visual Question Answering (VQA) requires models to not only extract accurate textual answers but also precisely localize them within document images, a capability critical for interpretability in high-stakes applications. However, existing systems achieve strong textual accuracy while producing unreliable spatial grounding, or sacrifice performance for interpretability. We present ARIAL (Agentic Reasoning for Interpretable Answer Localization), a modular framework that orchestrates specialized tools through an LLM-based planning agent to achieve both precise answer extraction and reliable spatial grounding. ARIAL decomposes Document VQA into structured subtasks: OCR-based text extraction with TrOCR, retrieval-augmented context selection using semantic search, answer generation via a fine-tuned Gemma 3-27B model, and explicit bounding-box localization through text-to-region alignment. This modular architecture produces transparent reasoning traces, enabling tool-level auditability and independent component optimization. We evaluate ARIAL on four benchmarks (DocVQA, FUNSD, CORD, and SROIE) using both textual accuracy (ANLS) and spatial precision (mAP at IoU 0.50 to 0.95). ARIAL achieves state-of-the-art results across all datasets: 88.7 ANLS and 50.1 mAP on DocVQA, 90.0 ANLS and 50.3 mAP on FUNSD, 85.5 ANLS and 60.2 mAP on CORD, and 93.1 ANLS on SROIE, surpassing the previous best method (DLaVA) by +2.8 ANLS and +3.9 mAP on DocVQA. Our work demonstrates how agentic orchestration of specialized tools can simultaneously improve performance and interpretability, providing a pathway toward trustworthy, explainable document AI systems.
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。