The Era of Agentic Organization: Learning to Organize with Language Models

TL;DR

This paper introduces AsyncThink, a novel reasoning paradigm for large language models that enables asynchronous, collaborative problem-solving, reducing latency and improving accuracy in complex tasks.

Contribution

The paper proposes AsyncThink, a new asynchronous reasoning framework with dynamic task assignment and reinforcement learning optimization, advancing collaborative AI capabilities.

Findings

28% lower inference latency compared to parallel thinking

Improved accuracy on mathematical reasoning tasks

Generalizes to unseen tasks without additional training

Abstract

We envision a new era of AI, termed agentic organization, where agents solve complex problems by working collaboratively and concurrently, enabling outcomes beyond individual intelligence. To realize this vision, we introduce asynchronous thinking (AsyncThink) as a new paradigm of reasoning with large language models, which organizes the internal thinking process into concurrently executable structures. Specifically, we propose a thinking protocol where an organizer dynamically assigns sub-queries to workers, merges intermediate knowledge, and produces coherent solutions. More importantly, the thinking structure in this protocol can be further optimized through reinforcement learning. Experiments demonstrate that AsyncThink achieves 28% lower inference latency compared to parallel thinking while improving accuracy on mathematical reasoning. Moreover, AsyncThink generalizes its learned asynchronous thinking capabilities, effectively tackling unseen tasks without additional training.