Tag
This paper studies how reasoning exchange among multiple AI agents can improve accuracy but also risk error propagation, proposing a runtime monitoring framework to prevent such propagation.