RecursiveMAS

RecursiveMAS is a novel framework developed by researchers from the University of Illinois Urbana-Champaign and Stanford University. It addresses the limitations of current multi-agent AI systems that rely on text-based communication, which leads to latency, increased token costs, and difficulties in training the entire system cohesively. RecursiveMAS enables agents to collaborate and transmit information through embedding space, resulting in significant efficiency and performance improvements across complex domains like code generation, medical reasoning, and search. It offers a scalable and cost-effective blueprint for custom multi-agent systems by allowing agents to co-evolve as a single integrated whole, inspired by recursive language models.