Room "Sala Seminari" - Abacus Building (U14)
Beyond Single-Agent Architectures: Orchestrating AI Communities for Reliable Software Engineering
Speaker
Prof. Davide Di Ruscio
University of L'Aquila
Abstract: Utilizing LLM-based agents in Software Engineering demands significant improvements in accuracy and the mitigation of inherent biases. While the scale of Large Language Models continues to grow, single-agent hallucinations appear to be a fundamental byproduct of generative inference mechanisms. Emerging research suggests that the path toward reliable performance lies not in larger models, but in multi-agent collaboration and debate. Much like human professional environments, software quality improves through task specialization and structured peer review. In the AI domain, however, these collaborative cycles are remarkably cost-effective and rapid.
This talk introduces the EU MOSAICO project (https://mosaico-project.eu/), which establishes a theoretical and technical framework for scaling large AI-agent communities. We will explore the MOSAICO platform’s capabilities in orchestration, governance, and quality assurance. Finally, we discuss how the platform integrates with existing development environments to facilitate human-in-the-loop decision-making and agent reuse.
Bio: Davide Di Ruscio is a Full Professor at the Department of Information Engineering Computer Science and Mathematics of the University of L’Aquila. His main research interests are related to several aspects of Software Engineering, Open Source Software, and Model-Driven Engineering (MDE), including domain-specific modelling languages, model transformation, model differencing, coupled evolution, and recommendation systems. He has published more than 200 papers in various journals, conferences and workshops on such topics.
Contact person for the seminar: rafael.penaloza@unimib.it