使用DPO隐式奖励引导语言模型

Bootstrapping Language Models with DPO Implicit Rewards

June 14, 2024

作者: Changyu Chen, Zichen Liu, Chao Du, Tianyu Pang, Qian Liu, Arunesh Sinha, Pradeep Varakantham, Min Lin

cs.AI

摘要

在大型语言模型（LLMs）中的人类对齐是一个活跃的研究领域。最近的一项开创性工作，即直接偏好优化（DPO），通过绕过强化学习从人类反馈中学习（RLHF）中的奖励学习阶段，极大地简化了这一过程。DPO 在训练后提供了一个隐式奖励模型。在这项工作中，我们做出了一个新颖的观察，即这个隐式奖励模型本身可以被用于自身引导方式进一步对齐 LLM。我们的方法是利用当前 LLM 模型的奖励来构建一个偏好数据集，然后在后续的 DPO 轮次中使用。我们还加入了一些改进，消除了回应长度的偏见，并提高了偏好数据集的质量以进一步改进我们的方法。我们的方法，命名为使用 DPO 隐式奖励的自我对齐（DICE），在对齐方面取得了巨大进展，并在 AlpacaEval 2 上表现优异，以 27.55% 的长度受控胜率击败了 GPT-4 Turbo，但仅使用了 80 亿参数且没有外部反馈。我们的代码可在 https://github.com/sail-sg/dice 找到。

English

Human alignment in large language models (LLMs) is an active area of research. A recent groundbreaking work, direct preference optimization (DPO), has greatly simplified the process from past work in reinforcement learning from human feedback (RLHF) by bypassing the reward learning stage in RLHF. DPO, after training, provides an implicit reward model. In this work, we make a novel observation that this implicit reward model can by itself be used in a bootstrapping fashion to further align the LLM. Our approach is to use the rewards from a current LLM model to construct a preference dataset, which is then used in subsequent DPO rounds. We incorporate refinements that debias the length of the responses and improve the quality of the preference dataset to further improve our approach. Our approach, named self-alignment with DPO ImpliCit rEwards (DICE), shows great improvements in alignment and achieves superior performance than Gemini Pro on AlpacaEval 2, reaching 27.55% length-controlled win rate against GPT-4 Turbo, but with only 8B parameters and no external feedback. Our code is available at https://github.com/sail-sg/dice.