几何驱动的 OOD 检测器即类增量学习器
Geometry-driven OOD Detectors Are Class-Incremental Learners
把每个任务的分类头改造成基于固定 ETF 锚点的几何空间，使分类头同时具备 IND 识别与 OOD 拒识能力，从而让 CIL 无需额外的任务推断或 logit 校准；这套「几何化检测头 + ArcFace 收紧类内」的做法可直接迁移到视觉分类器的开放世界增量部署与 OOD 拒识场景。
原文依据 replacing the learnable head with fixed Equiangular Tight Frame (ETF) anchors
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论类增量学习（CIL）旨在随时间获取新类别而不遗忘已有知识。近期方法利用预训练模型（PTM）来抑制遗忘，但它们大多优化特征提取器，而忽略了关键的分类头。本文提出一个简洁的观点：若每个任务都配备一个既能识别分布内（IND）类别、又能拒识分布外（OOD）输入的分类器，则 CIL 便自然成立——输入只被判定其为分布内的分类头接受，否则被拒识。在严格的理论与实证研究支持下，我们发现这种能力由类间分离性与类内紧致性所刻画；若缺乏这两点，标准线性头与余弦相似度头仍属闭集，无法产生可用的 OOD 信号。为此，我们提出 GOD（几何驱动的 OOD 检测器），通过将可学习分类头替换为固定的等角紧框架（ETF）锚点，在单一几何空间中统一 IND 识别与 OOD 拒识；ETF 损失强制类间分离，ArcFace 损失进一步收紧类内紧致性。为提升效率，我们进一步引入参数高效的混合架构与高效推理策略，从而同时降低参数占用与推理开销。在多种增量设置与数据集上的大量实验表明，GOD 取得了最优结果。[^1]：代码与数据集见补充材料。
英文原文摘要 1539 字符
Class-Incremental Learning (CIL) seeks to acquire new classes over time without erasing prior knowledge. While recent methods leverage pre-trained models (PTMs) to curb forgetting, they largely optimize the feature extractor and overlook the crucial classification head. In this work, we advance a simple view: if each task is equipped with a classifier that has the ability to both recognize in-distribution (IND) classes and reject out-of-distribution (OOD) inputs, CIL arises naturally--inputs are accepted only by heads that deem them in-distribution and rejected otherwise. Supported by rigorous theoretical and empirical studies, we find that this ability is characterized by Inter-class Separation and Intra-class Compactness; lacking these, standard linear and cosine-similarity heads remain closed-set and fail to yield a usable OOD signal. To address this, we propose GOD (Geometry-driven OOD Detectors), which unifies IND recognition and OOD rejection in a single geometric space by replacing the learnable head with fixed Equiangular Tight Frame (ETF) anchors; an ETF loss enforces inter-class separation, and an ArcFace loss further tightens intra-class compactness. For efficiency, we further introduce a parameter-efficient hybrid architecture and an efficient inference strategy, thus reducing both parameter footprint and inference cost. Extensive experiments on multiple incremental settings and datasets show that GOD achieves state-of-the-art results.[^1]: Code and datasets are available in the supplementary material.