Abstract

This paper tackles open-ended deep research (OEDR), a complex challenge whereAI agents must synthesize vast web-scale information into insightful reports.Current approaches are plagued by dual-fold limitations: static researchpipelines that decouple planning from evidence acquisition and one-shotgeneration paradigms that easily suffer from long-context failure issues like"loss in the middle" and hallucinations. To address these challenges, weintroduce WebWeaver, a novel dual-agent framework that emulates the humanresearch process. The planner operates in a dynamic cycle, iterativelyinterleaving evidence acquisition with outline optimization to produce acomprehensive, source-grounded outline linking to a memory bank of evidence.The writer then executes a hierarchical retrieval and writing process,composing the report section by section. By performing targeted retrieval ofonly the necessary evidence from the memory bank for each part, it effectivelymitigates long-context issues. Our framework establishes a new state-of-the-artacross major OEDR benchmarks, including DeepResearch Bench, DeepConsult, andDeepResearchGym. These results validate our human-centric, iterativemethodology, demonstrating that adaptive planning and focused synthesis arecrucial for producing high-quality, reliable, and well-structured reports.