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 vs 341 human ideas on 6 tasks. Advantage driven by novelty while maintaining usefulness. LLM teams benefit from efficient exploration (high semantic spread, shorter paths), while humans benefit from local conversational coherence.

Summary generated by Claude — human-verified