· ai
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, pe...