Hummer: Towards Limited Competitive Preference Dataset
This addresses a key challenge in aligning language models with human preferences for improved safety and adaptability, though it is incremental as it builds on existing datasets and methods.
The paper tackled the problem of conflicting alignment objectives in preference datasets, which increase vulnerability to jailbreak attacks, by introducing a novel statistical metric to quantify conflict and creating reduced-conflict datasets, resulting in reward models that effectively balance diverse objectives and reduce vulnerabilities.
Preference datasets are essential for incorporating human preferences into pre-trained language models, playing a key role in the success of Reinforcement Learning from Human Feedback. However, these datasets often demonstrate conflicting alignment objectives, leading to increased vulnerability to jailbreak attacks and challenges in adapting downstream tasks to prioritize specific alignment objectives without negatively impacting others. In this work, we introduce a novel statistical metric, Alignment Dimension Conflict, to quantify the degree of conflict within preference datasets. We then present \texttt{Hummer} and its fine-grained variant, \texttt{Hummer-F}, as innovative pairwise preference datasets with reduced-conflict alignment objectives. \texttt{Hummer} is built based on UltraFeedback and is enhanced by AI feedback from GPT-4, marking as the first preference dataset aimed at reducing the competition between alignment objectives. Furthermore, we develop reward models, HummerRM and HummerRM-F, which employ a hybrid sampling approach to balance diverse alignment objectives effectively. This sampling method positions HummerRM as an ideal model for domain-specific further fine-tuning and reducing vulnerabilities to attacks.