A paper by researchers of the Jožef Stefan Institute has been accepted at NeurIPS, the leading conference in artificial intelligence. Dr. Blaž Bertalanič and Dr. Carolina Fortuna of the Department of Communication Systems are the authors of "The Ringelmann Effect in Multi-Agent LLM Systems: A Scaling Law for Effective Team Size." NeurIPS is one of the most prestigious and selective AI conferences, where leading research groups present their work. Inspired by the Ringelmann effect, the observation that individuals contribute less effort as a group grows, the authors derived a scaling law describing how a group’s effective size changes with the nominal number of AI agents. Their study, covering 44 combinations of models, tasks and conditions, shows that larger groups of collaborating agents do not necessarily produce greater response diversity or better performance than a single agent. The point where adding agents stops helping can be predicted from a pilot with as few as five agents, helping developers avoid wasting computing resources.