Alleviating Hallucinations in Large Language Models with Skepticism Modeling

ACL ARR 2024 December Submission1619 Authors

Abstract: Hallucinations is a major challenge for large language models (LLMs), prevents adoption in diverse fields. Uncertainty estimation could be used for alleviating the damages of hallucinations. The skeptical emotion of human could be useful for enhancing the ability of self estimation. Inspirited by this observation, we proposed a new approach called Skepticism Modeling (SM), which is formalized by combining the information of tokens and probabilities for self estimation. We construct the doubt emotion aware data, perform continual pre-training, and then fine-tune the LLMs, improve their ability of self estimation. Experimental results demonstrate this new approach effectively enhances a model's ability to estimate their uncertainty, and validate its generalization ability of other tasks by out-of-domain experiments.

Paper Type: Long

Research Area: Language Modeling

Research Area Keywords: fine-tuning

Contribution Types: NLP engineering experiment

Languages Studied: English

Submission Number: 1619