Halo: Estimation And Reduction Of Hallucinations In Open-source Weak Large Language Models

Elaraby Mohamed, Lu Mengyin, Dunn Jacob, Zhang Xueying, Wang Yu, Liu Shizhu, Tian Pingchuan, Wang Yuping, Wang Yuxuan. Arxiv 2023

[Paper]    
Applications Tools

Large Language Models (LLMs) have revolutionized Natural Language Processing (NLP). Although convenient for research and practical applications, open-source LLMs with fewer parameters often suffer from severe hallucinations compared to their larger counterparts. This paper focuses on measuring and reducing hallucinations in BLOOM 7B, a representative of such weaker open-source LLMs that are publicly available for research and commercial applications. We introduce HaloCheck, a lightweight BlackBox knowledge-free framework designed to quantify the severity of hallucinations in LLMs. Additionally, we explore techniques like knowledge injection and teacher-student approaches to alleviate hallucinations in low-parameter LLMs. Our experiments effectively demonstrate the reduction of hallucinations in challenging domains for these LLMs.

Similar Work