Fine-grained control over large language models (LLMs) remains a significant challenge, hindering their adaptability to diverse user needs. While Reinforcement Learning from Human Feedback (RLHF) shows promise in aligning LLMs, its reliance on scalar rewards often limits its ability to capture diverse user preferences in real-world applications. To address this limitation, we introduce the Directional Preference Alignment (DPA) framework. Unlike the scalar-reward RLHF, DPA incorporates multi-objective reward modeling to represent diverse preference profiles. Additionally, DPA models user preferences as directions (i.e., unit vectors) in the reward space to achieve user-dependent preference control. Our method involves training a multi-objective reward model and then fine-tuning the LLM with a preference-conditioned variant of Rejection Sampling Finetuning (RSF), an RLHF method adopted by Llama 2. This method enjoys a better performance trade-off across various reward objectives. In comparison with the scalar-reward RLHF, DPA offers users intuitive control over LLM generation: they can arithmetically specify their desired trade-offs (e.g., more helpfulness with less verbosity). We also validate the effectiveness of DPA with real-world alignment experiments on Mistral-7B. Our method provides straightforward arithmetic control over the trade-off between helpfulness and verbosity while maintaining competitive performance with strong baselines such as Direct Preference Optimization (DPO).

翻译：大语言模型（LLM）的细粒度控制仍然是一个重大挑战，阻碍了其适应多样化用户需求的能力。尽管基于人类反馈的强化学习（RLHF）在对齐LLM方面展现了前景，但其对标量奖励的依赖往往限制了捕捉实际应用中多样化用户偏好的能力。为解决这一局限，我们提出了方向偏好对齐（DPA）框架。不同于标量奖励的RLHF，DPA引入多目标奖励建模以表征多样化的偏好分布。此外，DPA将用户偏好建模为奖励空间中的方向（即单位向量），从而实现依赖用户偏好的控制。我们的方法包括训练多目标奖励模型，随后使用偏好条件化的拒绝采样微调（RSF）变体对LLM进行微调——RSF是Llama 2采用的RLHF方法。该方法在不同奖励目标之间实现了更优的性能权衡。与标量奖励RLHF相比，DPA为用户提供了对LLM生成的直观控制：他们可以通过算术方式指定期望的权衡（例如，提升有用性的同时降低冗长度）。我们还通过在Mistral-7B上的实际对齐实验验证了DPA的有效性。我们的方法在保持与直接偏好优化（DPO）等强基线方法竞争性能的同时，提供了对有用性与冗长度之间权衡的直接算术控制。