AI models that simulate internal debate dramatically improve accuracy on complex tasks

A new study by Google suggests that advanced reasoning models achieve high performance by simulating multi-agent-like debates involving diverse perspectives, personality traits, and domain expertise.Their experiments demonstrate that this internal de...