Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Yunque DeepResearch Technical Report

Yuxuan Cai, Xinyi Lai, Peng Yuan, Weiting Liu, Huajian Li, Mingda Li, Xinghua Wang, Shengxie Zheng, Yanchao Hao, Yuyang Yin, Zheng Wei

TL;DR

Yunque DeepResearch addresses pivotal bottlenecks in autonomous deep-research agents—cognitive overload in long-horizon tasks, fragility leading to cascading errors, and limited modular extensibility. It presents a hierarchical, modular framework with a central Main Agent, Dynamic Context Management via sub-goal–driven structured memory, and a Supervisor for anomaly detection and context pruning, all supported by an Atomic Capability Pool of specialized sub-agents and basic tools. The approach demonstrates state-of-the-art performance on benchmarks such as GAIA, BrowseComp, BrowseComp-ZH, and Humanity's Last Exam, with ablations confirming the essential roles of memory, supervision, and specialized modules. The work also shows model-agnostic gains across backbones and contributes an open-source release to accelerate development of robust, collaborative agentic systems for complex information-seeking tasks.

Abstract

Deep research has emerged as a transformative capability for autonomous agents, empowering Large Language Models to navigate complex, open-ended tasks. However, realizing its full potential is hindered by critical limitations, including escalating contextual noise in long-horizon tasks, fragility leading to cascading errors, and a lack of modular extensibility. To address these challenges, we introduce Yunque DeepResearch, a hierarchical, modular, and robust framework. The architecture is characterized by three key components: (1) a centralized Multi-Agent Orchestration System that routes subtasks to an Atomic Capability Pool of tools and specialized sub-agents; (2) a Dynamic Context Management mechanism that structures completed sub-goals into semantic summaries to mitigate information overload; and (3) a proactive Supervisor Module that ensures resilience through active anomaly detection and context pruning. Yunque DeepResearch achieves state-of-the-art performance across a range of agentic deep research benchmarks, including GAIA, BrowseComp, BrowseComp-ZH, and Humanity's Last Exam. We open-source the framework, reproducible implementations, and application cases to empower the community.

Yunque DeepResearch Technical Report

TL;DR

Yunque DeepResearch addresses pivotal bottlenecks in autonomous deep-research agents—cognitive overload in long-horizon tasks, fragility leading to cascading errors, and limited modular extensibility. It presents a hierarchical, modular framework with a central Main Agent, Dynamic Context Management via sub-goal–driven structured memory, and a Supervisor for anomaly detection and context pruning, all supported by an Atomic Capability Pool of specialized sub-agents and basic tools. The approach demonstrates state-of-the-art performance on benchmarks such as GAIA, BrowseComp, BrowseComp-ZH, and Humanity's Last Exam, with ablations confirming the essential roles of memory, supervision, and specialized modules. The work also shows model-agnostic gains across backbones and contributes an open-source release to accelerate development of robust, collaborative agentic systems for complex information-seeking tasks.

Abstract

Deep research has emerged as a transformative capability for autonomous agents, empowering Large Language Models to navigate complex, open-ended tasks. However, realizing its full potential is hindered by critical limitations, including escalating contextual noise in long-horizon tasks, fragility leading to cascading errors, and a lack of modular extensibility. To address these challenges, we introduce Yunque DeepResearch, a hierarchical, modular, and robust framework. The architecture is characterized by three key components: (1) a centralized Multi-Agent Orchestration System that routes subtasks to an Atomic Capability Pool of tools and specialized sub-agents; (2) a Dynamic Context Management mechanism that structures completed sub-goals into semantic summaries to mitigate information overload; and (3) a proactive Supervisor Module that ensures resilience through active anomaly detection and context pruning. Yunque DeepResearch achieves state-of-the-art performance across a range of agentic deep research benchmarks, including GAIA, BrowseComp, BrowseComp-ZH, and Humanity's Last Exam. We open-source the framework, reproducible implementations, and application cases to empower the community.
Paper Structure (39 sections, 10 equations, 2 figures, 3 tables)

This paper contains 39 sections, 10 equations, 2 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. DeepResearch Agent
  4. Working Memory Management
  5. Specialized Agent
  6. Framework
  7. Main Agent
  8. Memory
  9. Structured Memory Generation
  10. Memory Unit Structure
  11. Dynamic Folding and Adding Mechanism
  12. Dynamic Context Management
  13. Adaptive Context Construction
  14. Efficiency and Complexity
  15. Atomic Capability Pool
  16. ...and 24 more sections

Figures (2)

  • Figure 1: Overview of Yunque DeepResearch.
  • Figure 2: Illustration of Data Analysis Agent.