AIBullisharXiv – CS AI · Mar 37/103
🧠
MAS-Orchestra: Understanding and Improving Multi-Agent Reasoning Through Holistic Orchestration and Controlled Benchmarks
Researchers introduce MAS-Orchestra, a new framework for multi-agent AI systems that uses reinforcement learning to orchestrate multiple AI agents more efficiently. The system achieves 10x efficiency improvements over existing methods and includes a benchmark (MASBENCH) to better understand when multi-agent systems outperform single-agent approaches.