高准确率的噪声标签学习模型可能无法识别异常输入，因误分类样本与异常数据在特征上重叠。
When Accuracy Is Not Enough: Uncertainty Collapse between Noisy Label Learning and Out-of-Distribution Detection
- 提出无模型依赖的ACC-OOD基准，统一评估噪声标签训练后的模型对异常输入的检测能力。
- 发现高闭集准确率模型在真实/合成噪声下常出现不确定性坍缩，导致异常检测失败。
- 引入虚拟边界正则化（VMR）轻量修复，提升异常检测能力但不牺牲闭集性能。
以噪声标签学习（LNL）通常仅以闭集分类准确率作为评价标准，但实际部署中需具备拒绝分布外（OOD）输入的能力。本文提出一种无需依赖学习器的ACC-OOD基准，冻结LNL检查点，在合成与真实噪声场景下，使用标准化的近/OOD路由与后验得分进行评估。结果显示：高闭集准确率并不保证OOD可靠性，因为低置信度的误分类样本会与异常输入在分数和特征空间中发生结构重叠，形成‘不确定性坍缩’。该重叠导致主流方法在标准OOD评分下丧失对异常输入的可分性。为应对此问题，研究了轻量级修复方法虚拟边界正则化（VMR），其通过在可信内分布批次上生成虚拟边界异常样本，扩大能量间隙。实验表明，VMR在不替换原目标函数且不降低闭集准确率的前提下，部分缓解了坍缩引发的远端OOD失败。结果支持构建同时报告闭集泛化、开放世界可靠性与结构重叠诊断的综合评测体系。
原文摘要 · Abstract (English)
Learning with noisy labels (LNL) is typically benchmarked by closed-set classification accuracy, yet deployment often requires classifiers to reject out-of-distribution (OOD) inputs. We present a learner-agnostic ACC-OOD benchmark that freezes LNL checkpoints and evaluates them with standardized near-/far-OOD routing and post-hoc scores across synthetic and real label noise. The benchmark reveals a recurring failure mode: high closed-set accuracy does not ensure OOD reliability, because low-confidence, misclassified in-distribution samples can overlap the score and feature regions occupied by OOD inputs under noisy training. We term this pathology uncertainty collapse. This structural overlap can make high-accuracy LNL methods lose separability at the ID-error/OOD interface under standard OOD scores. As an intervention, we study Virtual Margin Regularization (VMR), a lightweight repair probe demonstrated mainly with PSSCL that synthesizes boundary virtual outliers on trusted ID batches and widens the energy margin. VMR partially reduces the collapse-induced far-OOD failure without replacing the host objective or sacrificing closed-set accuracy in the tested settings. These results support LNL benchmarks that co-report closed-set generalization, open-world reliability, and structural overlap diagnostics.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。