φ-DPO：面向大型多模态模型持续学习的公平直接偏好优化方法
$\phi$-DPO: Fairness Direct Preference Optimization Approach to Continual Learning in Large Multimodal Models
提出基于直接偏好优化（DPO）的持续学习范式，并设计 φ-DPO 损失显式处理数据不平衡，可复用于多模态大模型在视觉任务上的持续学习与公平性提升。
原文依据 we first propose a new continual learning paradigm based on Direct Preference Optimization (DPO) to mitigate catastrophic forgetting
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论大型多模态模型（LMMs）持续学习中的公平性是一个新兴但尚未充分探索的挑战，尤其是在数据分布不平衡的情况下，这可能导致模型更新有偏且跨任务性能次优。尽管近期的持续学习研究在缓解灾难性遗忘方面取得了进展，但由数据不平衡引起的公平性问题仍很大程度上未被探索。本文提出了一种新颖的公平直接偏好优化（FaiDPO 或 φ-DPO）框架，用于 LMMs 的持续学习。具体而言，我们首先提出了一种基于直接偏好优化（DPO）的新持续学习范式，通过将学习与成对偏好信号对齐来缓解灾难性遗忘。然后，我们指出了传统 DPO 在不平衡数据下的局限性，并提出了一种新的 φ-DPO 损失，显式地处理分布偏差。我们提供了全面的理论分析，证明我们的方法同时解决了遗忘和数据不平衡问题。此外，为了支持基于 φ-DPO 的持续学习，我们为现有基准构建了持续学习场景下的成对偏好标注。大量实验和消融研究表明，所提出的 φ-DPO 在多个基准上达到了最优性能，优于先前的 LMMs 持续学习方法。
英文原文摘要 1401 字符
Fairness in Continual Learning for Large Multimodal Models (LMMs) is an emerging yet underexplored challenge, particularly in the presence of imbalanced data distributions that can lead to biased model updates and suboptimal performance across tasks. While recent continual learning studies have made progress in addressing catastrophic forgetting, the problem of fairness caused by imbalanced data remains largely underexplored. This paper presents a novel Fairness Direct Preference Optimization (FaiDPO or \phi-DPO) framework for continual learning in LMMs. In particular, we first propose a new continual learning paradigm based on Direct Preference Optimization (DPO) to mitigate catastrophic forgetting by aligning learning with pairwise preference signals. Then, we identify the limitations of conventional DPO in imbalanced data and present a new \phi-DPO loss that explicitly addresses distributional biases. We provide a comprehensive theoretical analysis demonstrating that our approach addresses both forgetting and data imbalance. Additionally, to enable \phi-DPO-based continual learning, we construct pairwise preference annotations for existing benchmarks in the context of continual learning. Extensive experiments and ablation studies show the proposed \phi-DPO achieves State-of-the-Art performance across multiple benchmarks, outperforming prior continual learning methods of LMMs.