NoVo: Norm Voting off Hallucinations with Attention Heads in Large Language Models
This addresses the critical issue of factual inaccuracies in LLMs for high-stakes applications, offering a scalable and generalizable solution that surpasses existing methods.
The paper tackles the problem of hallucinations in Large Language Models by introducing Norm Voting (NoVo), a lightweight method that uses attention head norms to improve factual accuracy in zero-shot multiple-choice questions, achieving at least 19 accuracy points improvement on TruthfulQA MC1 and significant gains on 90% of 20 diverse datasets.
Hallucinations in Large Language Models (LLMs) remain a major obstacle, particularly in high-stakes applications where factual accuracy is critical. While representation editing and reading methods have made strides in reducing hallucinations, their heavy reliance on specialised tools and training on in-domain samples, makes them difficult to scale and prone to overfitting. This limits their accuracy gains and generalizability to diverse datasets. This paper presents a lightweight method, Norm Voting (NoVo), which harnesses the untapped potential of attention head norms to dramatically enhance factual accuracy in zero-shot multiple-choice questions (MCQs). NoVo begins by automatically selecting truth-correlated head norms with an efficient, inference-only algorithm using only 30 random samples, allowing NoVo to effortlessly scale to diverse datasets. Afterwards, selected head norms are employed in a simple voting algorithm, which yields significant gains in prediction accuracy. On TruthfulQA MC1, NoVo surpasses the current state-of-the-art and all previous methods by an astounding margin -- at least 19 accuracy points. NoVo demonstrates exceptional generalization to 20 diverse datasets, with significant gains in over 90\% of them, far exceeding all current representation editing and reading methods. NoVo also reveals promising gains to finetuning strategies and building textual adversarial defence. NoVo's effectiveness with head norms opens new frontiers in LLM interpretability, robustness and reliability.