TAR：面向细粒度广义类别发现的词元感知精炼
TAR: Token-Aware Refinement for Fine-grained Generalized Category Discovery
指出 ViT 注意力在少数词元上异常高响应（注意力伪影）会压制细粒度局部线索，并提出用整个词元序列而非仅首个词元做分类的即插即用精炼模块，可直接移植到任何基于 ViT 特征的细粒度识别或开放类别发现流程中。
原文依据 attention artifacts where attention maps exhibit abnormally high responses on a few tokens
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论对于同时包含已知类别与未知类别的无标注数据集，广义类别发现（GCD）旨在对已知类别进行分类，同时发现未知类别。尽管现有方法在粗粒度数据集上取得了有前景的结果，但在细粒度场景中表现不佳。我们观察到，注意力伪影——即注意力图在少数词元上表现出异常高的响应——会通过使模型过度强调全局语义而忽略具有判别性的局部线索，显著干扰细粒度 GCD。为解决该问题，我们提出词元感知精炼（TAR）框架，这是一个可即插即用的模块，能够缓解注意力伪影并增强局部信息。与仅依赖首个词元进行分类的传统方法不同，TAR 利用整个词元序列以更好地捕捉细粒度细节。大量实验表明，TAR 在多个基准上均能稳定提升性能。我们的代码已发布于 https://github.com/VectorYangYiStar/TAR。
英文原文摘要 1049 字符
For an unlabeled dataset containing both known and unknown categories, Generalized Category Discovery (GCD) aims to classify known categories while discovering unknown ones. Although existing methods achieve promising results on coarse-grained datasets, they struggle in fine-grained scenarios. We observe that attention artifacts where attention maps exhibit abnormally high responses on a few tokens--significantly interfere with fine-grained GCD by causing the model to overemphasize global semantics and overlook discriminative local cues. To address this issue, we propose the Token-Aware Refinement (TAR) framework, a plug-and-play module that mitigates attention artifacts and enhances local information. Unlike conventional approaches that rely solely on the first token for classification, TAR leverages the entire token sequence to better capture fine-grained details. Extensive experiments demonstrate that TAR consistently improves performance across multiple benchmarks. Our code is available at https://github.com/VectorYangYiStar/TAR.