DocDancer: Towards Agentic Document-Grounded Information Seeking

2026-01-08

arXiv.org

AI Summary

This work addresses the limitations of existing document-based question answering agents, which typically rely on closed-source models and lack effective tool utilization capabilities, hindering efficient and open-ended access to document information. The study formulates document QA as an information-seeking task and proposes the first end-to-end trainable, open-source, tool-augmented agent framework that explicitly models the processes of document exploration and comprehension. To mitigate the scarcity of high-quality training data, the authors design an exploration-synthesis data generation pipeline and integrate a long-context understanding model to enhance performance. The approach achieves state-of-the-art results on the MMLongBench-Doc and DocBench benchmarks, demonstrating its effectiveness while offering novel insights into agent tool design and synthetic data construction for document understanding tasks.

Technology Category

• Natural Language Processing: Question Answering
• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Multiagent Systems: Modeling other Agents

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Document Question Answering (DocQA) focuses on answering questions grounded in given documents, yet existing DocQA agents lack effective tool utilization and largely rely on closed-source models. In this work, we introduce DocDancer, an end-to-end trained open-source Doc agent. We formulate DocQA as an information-seeking problem and propose a tool-driven agent framework that explicitly models document exploration and comprehension. To enable end-to-end training of such agents, we introduce an Exploration-then-Synthesis data synthesis pipeline that addresses the scarcity of high-quality training data for DocQA. Training on the synthesized data, the trained models on two long-context document understanding benchmarks, MMLongBench-Doc and DocBench, show their effectiveness. Further analysis provides valuable insights for the agentic tool design and synthetic data.

Problem

• Document Question Answering
• tool utilization
• open-source models
• training data scarcity
• agentic information seeking

Innovation

• DocDancer
• tool-driven agent
• document-grounded QA
• end-to-end training
• synthetic data pipeline