ARIAL：一个用于文档视觉问答的代理框架，具有精确答案定位
ARIAL: An Agentic Framework for Document VQA with Precise Answer Localization
AI总结：
ARIAL通过代理协调专门工具，实现文档VQA的高精度和可解释性，取得最佳性能和可解释性成果。
摘要：
文档视觉问答（VQA）要求模型不仅要提取准确的文本答案，还要在文档图像中精确定位它们，这种能力对于高风险应用中的可解释性至关重要。然而，现有系统在文本准确性方面表现优异，但空间定位不可靠，或者为了可解释性而牺牲性能。我们提出了ARIAL（用于可解释答案定位的代理推理），这是一个模块化的框架，通过基于大语言模型的规划代理协调专门的工具，以实现精确的答案提取和可靠的空域定位。ARIAL将文档VQA分解为结构化的子任务：使用TrOCR进行基于OCR的文本提取，使用语义搜索的检索增强上下文选择，通过微调的Gemma 3-27B模型进行答案生成，以及通过文本到区域对齐进行显式的边界框定位。这种模块化架构产生透明的推理轨迹，使工具层面的可审计性和独立组件的优化成为可能。我们在四个基准（DocVQA、FUNSD、CORD和SROIE）上评估了ARIAL，使用文本准确性（ANLS）和空间精度（mAP在IoU 0.50到0.95之间）。ARIAL在所有数据集上均取得最佳结果：在DocVQA上88.7 ANLS和50.1 mAP，在FUNSD上90.0 ANLS和50.3 mAP，在CORD上85.5 ANLS和60.2 mAP，在SROIE上93.1 ANLS，超过了先前最佳方法（DLaVA）在DocVQA上的+2.8 ANLS和+3.9 mAP。我们的工作展示了如何通过专门工具的代理协调同时提高性能和可解释性，为可信、可解释的文档AI系统提供了一条路径。
英文摘要：
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.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
