让高置信度对抗训练更可靠，避免被无关背景干扰。
Adaptive Causal Alignment for High-Confidence Adversarial Training
- 通过动态识别背景干扰，只修正有害的高置信预测。
- 在多个数据集上提升鲁棒性，比基线减少30%以上泛化差距。
- 适合关注模型可靠性与对抗训练稳定性的研究者。
逆向对抗训练利用高置信度预测来稳定鲁棒学习，但我们发现一个关键矛盾：高置信度常源于对非因果背景相关性的过拟合，而非物体本质语义。研究表明，视觉上下文具有双重属性，既可作为必要先验，也可能成为虚假混淆因子。现有盲目抑制策略因此存在缺陷，导致严重特征损失。为此，我们提出高置信度因果对齐训练（HICAT），构建语义均衡框架。其采用“测量-去偏-对齐”流程，集成可学习背景偏差估计器（LBBE）自适应诊断上下文效用，并通过自适应去偏机制进行精准逻辑值修正，辅以基于几何的前景逻辑正交增强（FLOE）损失，强制特征解耦。在CIFAR-10、CIFAR-100和ImageNet-1K上的大量实验表明，HICAT在多种架构（CNNs与ViTs）下均持续优于对应基线，显著缩小鲁棒泛化差距。
原文摘要 · Abstract (English)
Inverse adversarial training leverages high-confidence predictions to stabilize robust learning, yet we uncover a critical paradox: high confidence often stems from overfitting to non-causal background correlations rather than intrinsic object semantics. Our investigation reveals that visual context functions as a dual-natured signal, serving as either a necessary supportive prior or a spurious confounder. This insight renders existing blind suppression strategies flawed, as they inevitably lead to severe Feature Loss. To resolve this, we propose High-Confidence Causally Aligned Training (HICAT), a unified framework that establishes a Semantic Equilibrium. Operating on a ``Measure-Debias-Align'' pipeline, HICAT integrates a Learnable Background-Bias Estimator (LBBE) to adaptively diagnose context utility. Guided by this diagnosis, an Adaptive Debiasing mechanism performs surgical logit rectification, complemented by a geometrically grounded Foreground Logit Orthogonal Enhancement (FLOE) loss to enforce rigorous feature disentanglement. Extensive experiments on CIFAR-10, CIFAR-100, and ImageNet-1K demonstrate that HICAT consistently improves over matched baselines across diverse architectures (CNNs and ViTs) while significantly reducing the robust generalization gap.
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。