以扩展 Logit 归一化增强分布外检测
Enhancing Out-of-Distribution Detection with Extended Logit Normalization
ELogitNorm 是一个无超参的训练期正则目标，可直接替换交叉熵训练来提升任意 post-hoc OOD 打分函数（MSP、能量分数等）的效果，同时改善 ID 置信度校准；做视觉分类模型部署时可直接套用，无需改动推理流程。
原文依据 we introduce a feature distance-aware normalization objective, termed ELogitNorm
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论分布外（OOD）检测对于机器学习模型的安全部署至关重要。尽管已有大量工作致力于设计有效的 OOD 检测打分函数，但相对较少的研究探索以校准为导向的目标来训练神经网络，这类目标往往会损害预测精度并限制打分函数的选择。在本工作中，我们首先识别出 Logit 归一化（LogitNorm）中的特征坍缩问题，随后提出一种新颖的无超参数训练形式，可显著提升多种事后检测方法的效果。具体而言，我们引入一种特征距离感知的归一化目标，称为 ELogitNorm，它同时提升了 OOD 检测性能与分布内（ID）置信度校准。在标准基准上的大量实验表明，我们的方法在 OOD 检测上优于最优的训练期方法，同时保持了较强的 ID 分类性能。代码已公开于：https://github.com/limchaos/ElogitNorm。
英文原文摘要 1054 字符
Out-of-distribution (OOD) detection is essential for the safe deployment of machine learning models. While extensive work has focused on designing effective scoring functions for OOD detection, relatively few studies explore training neural networks with calibration-oriented objectives, which often compromise predictive accuracy and restrict the choice of scoring functions. In this work, we first identify feature collapse in Logit Normalization (LogitNorm), and then propose a novel hyperparameter-free training formulation that significantly improves a wide range of post-hoc detection methods. Specifically, we introduce a feature distance-aware normalization objective, termed ELogitNorm, which enhances both OOD detection performance and in-distribution (ID) confidence calibration. Extensive experiments on standard benchmarks demonstrate that our approach outperforms state-of-the-art training-time methods in OOD detection while preserving strong ID classification performance. Our code is available at: https://github.com/limchaos/ElogitNorm.