TFG/TFM: Multi-Agent Debate: A Framework for Collective Reasoning and Decision-Making

Motivation: The emergence of large language models (LLMs) and advanced multi-agent systems has opened new possibilities for automated reasoning and decision-making. Among these approaches, multi-agent debate stands out as a promising paradigm where multiple autonomous agents engage in structured discussions to reach more robust conclusions than a single model could achieve.

Objectives: The main objective of this work is to continue the development of a framework for multi-agent debate, grounded in a taxonomy that characterizes the different properties of such debates. This taxonomy provides a structured way to analyze and implement debate systems according to key dimensions such as agreement mechanisms, interaction format and topology, and participant roles. By extending this framework, the project seeks to enable systematic experimentation with various debate configurations, facilitating the evaluation of how specific properties affect the quality of reasoning, consensus formation, and overall system performance. In this work, the student will:
- Extend and refine the existing multi-agent debate framework by integrating new aspects of the proposed taxonomy.
- Design and implement diverse debate configurations of the taxonomy.
- Conduct systematic experiments to evaluate how these properties influence the quality, efficiency, and other relevant metrics of debates among agents.

Expected Results:
- An extended multi-agent debate framework
- The design and execution of a set of experiments of debates in Requirements Engineering tasks.
- Insights into how different properties of the debate affect system performance and outcomes.

Technologies:  
   Python, RESTful services.