Latent Agents: A Breakthrough in AI Technology for Internalizing Multi-Agent Debate
In the field of artificial intelligence, an innovative research called "Latent Agents" is quietly transforming the paradigm of multi-agent debate by internalizing the collaborative debate process into a single large language model (LLM), significantly reducing computational resource consumption while retaining the reasoning advantages of multi-agent systems. This breakthrough, achieved through two-stage fine-tuning, reduces token usage by up to 93% and has sparked heated discussions in the AI community, highlighting its efficiency and potential for secure applications.