发表机构
Institute of Information Engineering, Chinese Academy of Sciences; School of Cyber Security, University of Chinese Academy of Sciences; Nankai University; Tsinghua University; Harbin Institute of Technology（中国科学院信息工程研究所; 中国科学院大学网络空间安全学院; 南开大学; 清华大学; 哈尔滨工业大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
针对多标签类增量学习中已知与未知知识边界模糊的问题，提出KBK框架，通过分层特征纯化等模块缓解遗忘，在MS-COCO基准上超越最优方法2.7%。
AI 中文摘要
现有的类增量学习方法在多标签场景（MLCIL）中表现不佳，原因在于共现标签与不完整标签所产生的学习目标存在内在矛盾。我们认为核心障碍是模型在已知与未知知识之间的边界模糊，这会破坏历史知识保留、增加当前任务学习的复杂性，并限制对未来概念的适应性。为解决该问题，我们提出KBK（Knowing Beyond the Known），一种强化知识规范框架，该框架显式建模已知或未知内容，以统一历史、当前和未来学习。具体而言，为明确已知知识，我们开发了分层特征纯化模块，从全局特征中解耦细粒度的类特定特征，其中高级语义抽象通过低级视觉特征得到强化。此外，不确定性感知召回增强策略基于分布先验抑制不可靠预测，提升历史召回质量。为探测未知，KBK利用语义相关性在共现条件下合成信息丰富的未知特征，为未来学习保留嵌入空间。此外，为缓解异构遗忘，我们设计了类别平衡梯度补偿损失，根据遗忘速度动态调整梯度反向传播的权重。在多个基准上的实验验证了KBK的有效性和鲁棒性，在MS-COCO B0-C10设置下，即使不使用任何回放缓冲区，KBK的平均准确率（Avg. Acc）也超过了之前的最优方法2.7%。
英文摘要
Existing class-incremental learning methods struggle in multi-label scenarios (MLCIL) due to the inherent contradiction of learning objectives arising from co-occurring and incomplete labels. We argue that the core obstacle is the model's ambiguous boundary between known and unknown knowledge, which undermines historical knowledge retention, complicates current task learning, and limits adaptability to future concepts. To address this, we propose KBK (Knowing Beyond the Known), a reinforced knowledge specification framework that explicitly models what is known or not to unify historical, current, and prospective learning. Specifically, to clarify known knowledge, we develop a hierarchical feature purification module that disentangles fine-grained class-specific features from global features, where high-level semantic abstraction is reinforced with low-level visual features. Additionally, an uncertainty-aware recall enhancement strategy suppresses unreliable predictions based on distribution priors, improving the quality of historical recall. For probing the unknown, KBK leverages semantic correlations to synthesize informative unknown features under co-occurring, preserving embedding space for future learning. Furthermore, to mitigate heterogeneous forgetting, we design a category-balanced gradient compensation loss that dynamically reweights gradient backpropagation according to forgetting speeds. Experiments on multiple benchmarks validate the effectiveness and robustness of KBK, which surpasses prior best methods by 2.7% in Avg. Acc on MS-COCO B0-C10 setting even without any replay buffers.
