doi: 10.7763/IJET.2026.V18.1350
DeepResearch: A Survey of LLM-based Research Agents
- 1 School of Information Science and Technology, Yunnan Normal University, Kunming, China
- 2 AI Products Department, China National Knowledge Infrastructure, Beijing, China
- Manuscript received November 14, 2025
- accepted December 1, 2025
- published June 30, 2026
Abstract
Large Language Models (LLMs) have demonstrated remarkable advances in natural language processing and contentgeneration, yet they remain limited in supporting complexresearch tasks due to hallucinations, shallow summarization,weak multi-step reasoning, and unverifiable outputs. To addressthese challenges, the concept of DeepResearch has recentlyemerged, referring to research-oriented agents built on longcontextLLMs and augmented with capabilities such as multistepreasoning, deep retrieval, grounding and citation, agenticorchestration, and structured generation. These mechanismsenable models to act as autonomous researchers, capable ofplanning, retrieving, and producing systematic, evidence-basedreports. Representative systems include Google Gemini DeepResearch and OpenAI Deep Research, while in China platformssuch as Kimi, Baidu Wenxin, Doubao, and DeepSeek are activelydeveloping localized solutions. This paper provides a taxonomyof core techniques, presents a layered system architecture andarchitectural paradigms, reviews representative implementationsand applications, and highlights open challenges and futuredirections. By consolidating current progress, we aim to guidethe development of reliable and trustworthy DeepResearch agents.
Keywords
- DeepResearch;
- Large
- Language
- Models;
- ResearchAgents;
- Retrieval-Augmented
- Generation;
- Agentic
How to Cite
Jinyan Cai and Ruochong Yao, "DeepResearch: A Survey of LLM-based Research Agents," International Journal of Engineering and Technology, vol. 18, no. 2, pp. 92-101, 2026. https://doi.org/ 10.7763/IJET.2026.V18.1350
Copyright & License
Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).