Improving Factuality and Reasoning in Language Models through Multiagent Debate
This addresses the issue of fallacious answers and hallucinations in language models for users relying on accurate AI-generated content, representing a novel method rather than an incremental improvement.
The paper tackles the problem of improving factuality and reasoning in large language models by introducing a multiagent debate approach where multiple model instances propose and debate responses over multiple rounds, resulting in significant enhancements in mathematical and strategic reasoning tasks and reduced hallucinations.
Large language models (LLMs) have demonstrated remarkable capabilities in language generation, understanding, and few-shot learning in recent years. An extensive body of work has explored how their performance may be further improved through the tools of prompting, ranging from verification, self-consistency, or intermediate scratchpads. In this paper, we present a complementary approach to improve language responses where multiple language model instances propose and debate their individual responses and reasoning processes over multiple rounds to arrive at a common final answer. Our findings indicate that this approach significantly enhances mathematical and strategic reasoning across a number of tasks. We also demonstrate that our approach improves the factual validity of generated content, reducing fallacious answers and hallucinations that contemporary models are prone to. Our approach may be directly applied to existing black-box models and uses identical procedure and prompts for all tasks we investigate. Overall, our findings suggest that such "society of minds" approach has the potential to significantly advance the capabilities of LLMs and pave the way for further breakthroughs in language generation and understanding.