Multi-expert Prompting Improves Reliability, Safety, and Usefulness of Large Language Models
This addresses the need for more trustworthy and effective LLM outputs for users, though it is incremental as it builds on prior ExpertPrompting work.
The paper tackles the problem of improving the reliability, safety, and usefulness of large language model (LLM) generation by introducing Multi-expert Prompting, which simulates multiple experts and aggregates their responses, resulting in state-of-the-art truthfulness with an 8.69% improvement over the best baseline using ChatGPT.
We present Multi-expert Prompting, a novel enhancement of ExpertPrompting (Xu et al., 2023), designed to improve the large language model (LLM) generation. Specifically, it guides an LLM to fulfill an input instruction by simulating multiple experts, aggregating their responses, and selecting the best among individual and aggregated responses. This process is performed in a single chain of thoughts through our seven carefully designed subtasks derived from the Nominal Group Technique (Ven and Delbecq, 1974), a well-established decision-making framework. Our evaluations demonstrate that Multi-expert Prompting significantly outperforms ExpertPrompting and comparable baselines in enhancing the truthfulness, factuality, informativeness, and usefulness of responses while reducing toxicity and hurtfulness. It further achieves state-of-the-art truthfulness by outperforming the best baseline by 8.69% with ChatGPT. Multi-expert Prompting is efficient, explainable, and highly adaptable to diverse scenarios, eliminating the need for manual prompt construction.