Research

A research team at DTU officially unveiled AgenTex v1.0, a multi-agent AI system for scientific research, on the morning of May 17, 2026.

Minh Tue, student and creator of the DTU AgenTex project, introduces the software. Photo by contributors

The software applies multi-agent AI architecture to assist with scientific research. It is designed to support the entire research workflow, enhance the quality of international publications, and foster innovation in academic environments.

Addressing key research bottlenecks

According to SCImago 2024, Vietnam currently ranks 49th globally in annual scientific publication output, but a considerable gap still remains with several countries in the region, such as Thailand, Malaysia, and Singapore.

Surveys conducted at research and educational institutions in Central Vietnam identified five common bottlenecks that hinder scientific research and publication:

-    IMRaD mistakes,
-    lack of review before submission,
-    complex LaTeX formatting (67% of researchers spend over 20% of their time on formatting, according to Overleaf 2023),
-    literature reviews taking from two to five days,
-    fragmented research-support tools.

In response to these challenges, DTU developed AgenTex as a technical solution envisioned as a virtual research assistant for scientists. It is aimed at optimizing technical tasks so researchers can devote more time to academic thinking and making valuable scientific contributions.

The system is built around the core human-in-the-loop philosophy, in which:

-    AI serves as a tool for assistance, suggestion, critique, and optimization of technical tasks in research;
-    researchers remain at the center and retain full decision-making authority throughout the research workflow through interaction with AI.

“AgenTex does not write on behalf of researchers, but acts as an intelligent companion,” Phan Vu Minh Tue, student and creator of DTU’s AgenTex project, explains. “It sharpens ideas, accelerates literature reviews, improves writing quality, reviews more carefully, and ensures compliance with international publication standards - to allow researchers to focus on what cannot be replaced: scientific thinking and contributing to knowledge.”

Multi-agent AI ecosystem

AgenTex v1.0 is built on a multi-agent AI architecture consisting of five system layers and an ecosystem of nine specialized AI agents operating synchronously through a single web interface.
Each agent performs a distinct role in the research cycle: generating ideas, literature synthesis, draft development, review, checking consistency, and standardization of LaTeX formatting.

AgenTex v1.0 interface. Photo by contributors

AgenTex v1.0 offers three core features:

-    Drafting: simultaneous search of ArXiv, Semantic Scholar, OpenAlex, and Tavily to generate IMRaD outlines, citation-supported abstracts, and LaTeX code. About 10–15 minutes for 15-20 references.
-    Reviewing: simulates peer review and checks arguments, data consistency, journal template compliance, and research ethics.
-    Formatting: compiles LaTeX directly in the browser; supports IEEE, Elsevier, Springer, and ACM templates; exports PDF, TeX, and ZIP.

The system enables simultaneous data retrieval from ArXiv, Semantic Scholar, OpenAlex, and Tavily and supports IMRaD-structure research outlines, peer-review simulation, and manuscript formatting according to IEEE, Elsevier, Springer, and ACM standards.

Leaders from the Vietnamese Ministry of Education & Training and DTU with the research team. Photo provided by DTU

The initial pilot study involved more than fifty lecturers and PhD students and studied their research and project development.

“Just like the PhD students and other students who tested the software, I believe AgenTex is more than a drafting tool; it’s a smart ecosystem that supports research and academic editing,” Dr Hoang Thi Huong, acting Provost of the DTU School of Languages, Humanities, & Social Sciences, explains.

“The system effectively assists with formatting, structure, and research-content organization. This gives researchers more time for deeper academic thinking and the development of core scientific values.”

Toward large-scale AI research platform

Following the trial launch of AgenTex v1.0, the research team plans to expand the platform during 2026 -2027 by integrating multimodal capabilities, collaborative multi-author workspaces, and cloud-service models for universities.
With the vision of becoming an “AI university” in the near future, DTU is gradually developing a technology ecosystem that integrates artificial intelligence across all university operations, from administration and training to learning, research, and entrepreneurship.

Mr Nguyen Son Hai, Deputy Director of the Department of Science, Technology, & Information

(Ministry of Education & Training). Photo by contributors

Mr Nguyen Son Hai, Deputy Director of the Department of Science, Technology, & Information (Ministry of Education & Training), attended the launch ceremony. “The launch of AgenTex v1.0 demonstrates that DTU has chosen a practical direction closely aligned with science & technology research in higher education - a field facing increasingly demanding requirements,” he remarked.

“In only about eight months, the research team successfully and effectively developed a first version of their AI agent. Of note is the decision to position AI as a support tool rather than a replacement for academic activities and research through the human-in-the-loop model, which is the right choice in the current context.

“The application of technology must also remain grounded in respect for the core values and the ethics of research, thereby contributing to a safe environment of AI application and laying the foundation for new breakthroughs in digital higher education.”

(Media Center)