About this talk
This talk, presented by Massimo Tisi from IMT Atlantique, delves into the MOSAICO project, which focuses on enhancing the reliability of AI systems through multi-agent architectures and orchestration strategies. The speaker discusses the limitations of single-agent approaches in handling complex specifications and how a multi-agent framework can address these challenges by decomposing tasks among specialized agents for generation, validation, and evaluation. The session highlights the hierarchical architecture of the MOSAICO system, which includes various agent types, such as collaboration and supervision agents, that work together to ensure effective task execution and consensus in evaluations. The architecture promotes explicit collaboration, agent reputation, and interoperability, making significant strides in the reliability of AI in software engineering.
