Agents2h ago
Multi-Agent Teams Often Underperform Single Experts
Research from Apple reveals that self-organizing LLM teams often fail to outperform individual experts. While developers assume autonomous collaboration creates synergy, these unconstrained interactions frequently hinder performance. The study suggests that fixed workflows remain superior to emergent coordination. Practitioners should prioritize structured agent roles over open-ended collaboration to maintain high accuracy.