Deep Research Agents: A Systematic Examination And Roadmap

2025-06-22

AI Summary

Existing deep research agents (DRAs) lack systematic architectural principles and suffer from fundamental limitations in benchmarking—namely, insufficient knowledge coverage, poor execution efficiency, and weak goal alignment. Method: We propose a dual-dimensional workflow taxonomy (static vs. dynamic), introduce a novel architecture classification centered on planning strategies and agent composition, and design a unified technical stack integrating API/browser-based retrieval, a modular tool framework (supporting code execution and multimodal processing), Model Context Protocol (MCP), multi-agent coordination, and iterative reflection mechanisms. Contribution/Results: We identify critical gaps in current evaluation benchmarks and release an open-source, continuously updated repository featuring a canonical technology roadmap and concrete recommendations for benchmark improvement. This work establishes foundational infrastructure for reproducible, rigorous, and scalable DRA research and development.

Technology Category

• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Multiagent Systems: Agent/AI Theories and Architectures
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Architectures and workflows that use LLMs for crowd work

Abstract

The rapid progress of Large Language Models (LLMs) has given rise to a new category of autonomous AI systems, referred to as Deep Research (DR) agents. These agents are designed to tackle complex, multi-turn informational research tasks by leveraging a combination of dynamic reasoning, adaptive long-horizon planning, multi-hop information retrieval, iterative tool use, and the generation of structured analytical reports. In this paper, we conduct a detailed analysis of the foundational technologies and architectural components that constitute Deep Research agents. We begin by reviewing information acquisition strategies, contrasting API-based retrieval methods with browser-based exploration. We then examine modular tool-use frameworks, including code execution, multimodal input processing, and the integration of Model Context Protocols (MCPs) to support extensibility and ecosystem development. To systematize existing approaches, we propose a taxonomy that differentiates between static and dynamic workflows, and we classify agent architectures based on planning strategies and agent composition, including single-agent and multi-agent configurations. We also provide a critical evaluation of current benchmarks, highlighting key limitations such as restricted access to external knowledge, sequential execution inefficiencies, and misalignment between evaluation metrics and the practical objectives of DR agents. Finally, we outline open challenges and promising directions for future research. A curated and continuously updated repository of DR agent research is available at: {https://github.com/ai-agents-2030/awesome-deep-research-agent}.

Problem

• Analyze foundational technologies of Deep Research agents
• Systematize approaches with taxonomy for agent architectures
• Evaluate current benchmarks and highlight key limitations

Innovation

• Dynamic reasoning and adaptive long-horizon planning
• Multi-hop information retrieval and iterative tool use
• Modular frameworks with Model Context Protocols (MCPs)