SAME：面向有限数据下异构增量学习的稀疏锚定模型编辑
SAME: Sparse and Anchored Model Editing for Heterogeneous Incremental Learning under Limited Data
把模型编辑（稀疏参数更新 + 键值知识对）当作持续学习的载体，用双重锚框约束（特征对齐锚 + 参数幅度锚）在少样本、任务边界未知的跨域序列上稳定注入新知识，可直接迁移到视觉语言模型的增量适配与免遗忘微调。
原文依据 SAME sparsely and selectively updates task-relevant model parameters to extract compact, task-specific key-value knowledge pairs from limited data.
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论现有的增量学习（IL）方法主要在两种设定下评估：单域类增量设定，或任务标识已知的多域任务增量设定。然而这些假设在现实应用中往往不成立。为弥补这一差距，我们提出异构增量学习（HIL），这是一个用于在真实且具有挑战性的条件下评估增量学习方法的新设定，其中任务边界模糊或未知、类分布在跨环境时动态变化、且训练数据有限。模型编辑天然适合这一具有挑战性的 HIL，因为它能够在保留模型能力的同时高效整合新知识。因此，我们提出一种新的稀疏锚定模型编辑（SAME）来应对 HIL。具体而言，SAME 稀疏且有选择地更新与任务相关的模型参数，以从有限数据中提取紧凑的、任务特定的键值知识对。利用这些任务知识对，模型在双重锚定约束下为新任务执行知识注入。知识锚使更新后的模型特征与原始模型特征对齐，参数锚则约束参数幅度，从而确保知识注入的稳定与一致。我们的方法仅需少量带标注样本即可高效求解 HIL，且不引入额外的模型参数。在 11 个多样化视觉语言数据集、22 个序列任务上的大量实验表明，我们的方法在平均准确率上比现有持续学习方法高出 6.8%，同时保留了 oracle 模型 95.8% 的性能，展现出很强的稳定性与跨域泛化能力。
英文原文摘要 1705 字符
Existing Incremental Learning (IL) methods are primarily evaluated under either a single-domain class-incremental setting, or a multi-domain task-incremental setting with known task identifiers. However, these assumptions often fail to hold in real-world applications. To bridge this gap, we introduce Heterogeneous Incremental Learning (HIL), a new setting for evaluating IL methods under realistic and challenging conditions, where task boundaries are ambiguous or unknown, class distributions shift dynamically across environments, and training data is limited. Model editing is inherently well-suited for this challenging HIL, as it allows for the efficient integration of new knowledge while preserving model capabilities. Thus, we propose a novel Sparse and Anchored Model Editing (SAME) for addressing HIL. Specifically, SAME sparsely and selectively updates task-relevant model parameters to extract compact, task-specific key-value knowledge pairs from limited data. Using these task knowledge pairs, the model performs knowledge injection for new tasks under double-anchor constraints. The knowledge anchor aligns the updated and original model features, while the parameter anchor constrains parameter magnitudes, ensuring stable and consistent knowledge injection. Our method can efficiently solve HIL using only a few labeled examples, without introducing additional model parameters. Extensive experiments on 11 diverse visual-language datasets across 22 sequential tasks show that our method outperforms existing continual learning approaches by 6.8% in average accuracy, while retaining 95.8% of the oracle model performance, demonstrating strong stability and cross-domain generalization.