Recursive agents collaborate through latent states

A publication about a paper describes recursive multi-agent systems in which agents refine latent thoughts and exchange internal states, rather than decoding text at every step. Text is generated only at the end of the process.

According to the publication, experiments found more precise collaboration, faster execution, and lower token use. The reported results do not show that the technique works for every model or task. If you use AI to study or apply this approach, avoid entering personal data or confidential information without authorization, and review outputs before using them.