Empowering Language Models with Active Inquiry for Deeper Understanding

• DOI: 10.48550/arxiv.2402.03719
• Publication Date: 2024-02-06
• Authors: Jing-Cheng Pang; Heng-Bo Fan; Pengyuan Wang; Jia‐Hao Xiao; Nan Tang; Si-Hang Yang; Chengxing Jia; Sheng-Jun Huang; Yang Yu

The rise of large language models (LLMs) has revolutionized the way that we interact with artificial intelligence systems through natural language. However, LLMs often misinterpret user queries because their uncertain intention, leading to less helpful responses. In human interactions, clarification is sought targeted questioning uncover obscure information. Thus, in this paper, introduce LaMAI (Language Model Active Inquiry), designed endow same level interactive engagement. leverages active learning techniques raise most informative questions, fostering a dynamic bidirectional dialogue. This approach not only narrows contextual gap but also refines output LLMs, aligning it more closely expectations. Our empirical studies, across variety complex datasets where have limited conversational context, demonstrate effectiveness LaMAI. method improves answer accuracy from 31.9% 50.9%, outperforming other question-answering frameworks. Moreover, scenarios involving participants, consistently generates responses are superior or comparable baseline methods than 82% cases. applicability further evidenced by its successful integration various highlighting potential for future models.