Multi-agent AI systems outperform human teams in creativity

In three linesMulti-agent LLM teams outperform human teams in creativity (Cohen's d=1.50) across 4,541 AI ideas versus 341 human ideas on six tasks. Advantage driven by novelty while maintaining usefulness. LLM teams benefit from efficient exploration (high semantic spread, shorter paths), while humans benefit from smooth conversational flow (high local coherence, frequent pivots).

Summary generated by Claude — human-verified