Improving Factuality in Large Language Models via Decoding-Time Hallucinatory and Truthful Comparators
This addresses the issue of unreliable outputs for users relying on LLMs for factual information, though it is an incremental improvement over existing methods.
The paper tackles the problem of factual inaccuracies (hallucinations) in Large Language Models by introducing a decoding-time framework that uses comparators to constrain token predictions, resulting in significant improvements in model performance and factuality across multiple tasks.
Despite their remarkable capabilities, Large Language Models (LLMs) are prone to generate responses that contradict verifiable facts, i.e., unfaithful hallucination content. Existing efforts generally focus on optimizing model parameters or editing semantic representations, which compromise the internal factual knowledge of target LLMs. In addition, hallucinations typically exhibit multifaceted patterns in downstream tasks, limiting the model's holistic performance across tasks. In this paper, we propose a Comparator-driven Decoding-Time (CDT) framework to alleviate the response hallucination. Firstly, we construct hallucinatory and truthful comparators with multi-task fine-tuning samples. In this case, we present an instruction prototype-guided mixture of experts strategy to enhance the ability of the corresponding comparators to capture different hallucination or truthfulness patterns in distinct task instructions. CDT constrains next-token predictions to factuality-robust distributions by contrasting the logit differences between the target LLMs and these comparators. Systematic experiments on multiple downstream tasks show that our framework can significantly improve the model performance and response factuality.