# Mitigating Language Model Hallucination with Interactive Question-Knowledge Alignment

**Authors:** Shuo Zhang, Liangming Pan, Junzhou Zhao, William Yang Wang  
**arXiv:** [2305.13669](https://arxiv.org/abs/2305.13669)  
**Date:** 2023-05-23  
**Summary page:** https://www.summarizepaper.com/en/paper/2305.13669v1/

## Key points

- Addressing language model hallucination through a novel framework
- Emphasizing the importance of question-knowledge alignment in reducing hallucination
- Showcasing superior performance in reducing hallucination and improving response accuracy
- Discussing related work on language model hallucination and clarification question generation techniques
- Contributing valuable insights into mitigating language model hallucination through an interactive framework combining machine intelligence with human input

## Summary

In this study, we address the persistent issue of language model hallucination through our novel framework , which effectively enhances question-knowledge alignment by leveraging both machine intelligence and human input. Our research highlights the significance of question-knowledge misalignment as a critical factor contributing to language model hallucination in real-world scenarios. Through comprehensive evaluations, we showcase the superior performance of  in reducing hallucination and improving the accuracy of generated responses. Furthermore, we discuss related work on language model hallucination and clarification question generation techniques. Overall, our study contributes valuable insights into mitigating language model hallucination through an interactive framework that combines  with .