CUE：概念感知的多标签扩展以缓解长尾学习中的概念混淆
CUE: Concept-Aware Multi-Label Expansion to Mitigate Concept Confusion in Long-Tailed Learning
把单标签监督换成由 CLIP 零样本视觉线索与 LLM 类级语义线索构成的多标签概念监督，并用加权 Logit Adjustment 辅助损失注入，可直接迁移到任何长尾识别/分类微调流程中缓解头类主导与类间判别性被破坏的问题。
原文依据 we propose CUE, Concept-aware Multi-label Expansion, which introduces multi-label concept signals to preserve disrupted inter-class relationships
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论长尾分布在真实世界识别任务中十分常见，其中少数头类拥有大量样本，而大多数尾类样本极少。近年来，由于基础模型表现优异，针对长尾学习微调基础模型受到关注。然而，现有方法大多只关注缓解长尾分布偏差，而忽视了长尾分布引起的概念混淆。本文研究该问题，并将其归因于长尾分布下单标签监督的互斥性：这种互斥性抑制了相关类别之间的特征共享，放大了头类的主导地位，导致类间判别性被破坏。为解决该问题，我们提出 CUE（概念感知的多标签扩展），引入多标签概念信号以保留被破坏的类间关系。具体而言，CUE 通过以下方式构建概念集：(i) 从零样本 CLIP 中提取实例级视觉线索；(ii) 利用大语言模型生成类级语义线索；这两类线索通过分别加权的二元 Logit 调整（BLA）辅助损失引入，并与基线 Logit 调整（LA）损失联合优化。在多个长尾基准上的实验表明，CUE 取得了均衡且强劲的性能，超越了近期最优方法。代码见：https://github.com/zhangruichi/CUE。
英文原文摘要 1416 字符
Long-tailed distributions are common in real-world recognition tasks, where a few head classes have many samples while most tail classes have very few. Recently, fine-tuning foundation models for long-tailed learning has gained attention due to their excellent performance. However, most existing methods focus solely on mitigating long-tailed distribution bias while overlooking concept confusion caused by the long-tailed distribution. In this paper, we study this problem and attribute it to the mutual exclusivity of single-label supervision under long-tailed distributions, which suppresses feature sharing among related classes and amplifies the dominance of head classes, leading to disrupted inter-class discriminality. To address this, we propose CUE, Concept-aware Multi-label Expansion, which introduces multi-label concept signals to preserve disrupted inter-class relationships. Specifically, CUE constructs concept sets by (i) extracting instance-level visual cues from zero-shot CLIP and (ii) generating class-level semantic cues with LLM; the two cues are incorporated via separately weighted Binary Logit-Adjustmen (BLA) auxiliary losses and jointly optimized with the baseline Logit-Adjustmen (LA) loss. Experiments on several long-tailed benchmarks, CUE achieves balanced and strong performance, surpassing recent state-of-the-art methods. Code is available at: https://github.com/zhangruichi/CUE.