Denario is a groundbreaking AI-powered tool poised to reshape the landscape of scientific research. Developed collaboratively by an international team of scientists from institutions such as the Flatiron Institute, Cambridge University, the Autonomous University of Barcelona, and the Institute of Cosmos Sciences of the University of Barcelona (ICCUB), Denario harnesses the capabilities of large language models to support researchers across every stage of the scientific process. From the initial spark of a hypothesis to the final interpretation of results, Denario offers a dynamic and modular system designed to enhance productivity and creativity in science.
At ICCUB, researchers Raúl Jiménez Tellado (ICREA-ICCUB), Pedro Tarancón (ICCUB), and Licia Verde (ICREA-ICCUB) played pivotal roles in both the technical development and conceptual framing of Denario. Jiménez and Tarancón demonstrated the tool’s potential by applying it to solve intricate problems in mathematical physics—an area where traditional AI systems often falter. Their work showcased Denario’s ability to engage with high-level theoretical challenges, pushing the boundaries of what AI can contribute to fundamental science.
Meanwhile, Licia Verde focused on the ethical dimensions of AI-assisted research. She helped shape the principles guiding Denario’s use, particularly in relation to transparency, accountability, and the evolving norms of scientific publishing. Her contributions have been instrumental in ensuring that Denario not only accelerates research but also aligns with responsible and rigorous scientific standards.
Denario itself is structured as a multi-agent system, with each AI agent specialized in a distinct task. These agents collaborate to assist with idea generation, literature review, data analysis, coding, debugging, and even manuscript writing. The system is built using AG2 and LangGraph frameworks, with cmbagent serving as the backend for research analysis. Users can interact with Denario through a command-line interface or a graphical user interface (DenarioApp), making it accessible to both technical and non-technical users.
One of Denario’s most innovative features is its ability to generate complete research papers, including figures and LaTeX-formatted manuscripts tailored to specific journal styles. For example, users can specify the APS (Physical Review) format and receive a ready-to-submit draft. The tool also allows manual input at any stage, enabling researchers to refine ideas, methods, or results with their own expertise.
Despite its capabilities, Denario is not intended to replace scientists. Its creators emphasize that it functions as an assistant, not an autonomous researcher. Human oversight remains essential, especially given that only a fraction of Denario’s outputs currently yield novel insights, and some results may include fabricated data. The tool’s value lies in its ability to streamline workflows, surface unexplored ideas, and foster interdisciplinary collaboration.
The ICCUB team’s involvement underscores the institute’s commitment to advancing both the technological and ethical frontiers of AI in science. Their work with Denario exemplifies how artificial intelligence can be integrated into research in a way that respects the integrity and creativity of scientific inquiry.
For further details, you can explore the Simons Foundation article or visit the Denario GitHub repository.
