Multi-agent teams might not be better than a single good model: Apple and Stanford paper

Researchers find that when left to their own devices, more agents are not always the answer.