BiasFlow: Geometric Monitoring and Backbone Regularization for Spurious Feature Reliance
Worst-group accuracy (WGA) evaluates a trained predictor but does not characterize how its frozen backbone behaves when a new head is learned. We introduce BiasFlow, a hook-based toolkit for monitoring class-attribute centroid alignment (IBMI), within-class centroid separation (W-IBMI), and feature-projection sensitivity. IBMI is confounded by class-attribute correlation and is not a measure of causal feature reliance. We pair these diagnostics with BiasFlow Regularization (BFR), a supervised, composable class-conditional centroid-alignment penalty. W-IBMI verifies the quantity BFR optimizes; it is scale dependent and does not independently establish attribute removal. Across the reported small-scale benchmarks, adding BFR improves or preserves mean WGA, with gains up to +26.0 pp on UrbanC
提出 BiasFlow 工具和正则化方法以监测和减少预测模型对虚假特征的依赖。
- 核心方法
- 引入了 BiasFlow 工具,通过监测类-属性质心对齐(IBMI)、类内质心分离(W-IBMI)和特征投影敏感度来评估模型。同时提出了 BiasFlow 正则化(BFR)方法,添加了一个监督的、可组合的类条件质心对齐惩罚项,以改善模型行为。
- 适合谁读
- 适合关注模型公平性和鲁棒性的研究者和工程师阅读。
- 要解决的问题
- 论文解决的问题是当学习新的预测头时,如何监测和减少预训练模型主干网络对虚假特征的依赖,进而提高最差群体准确率(WGA)。
- 关键实验
- 在小规模基准测试中(如 UrbanC 数据集),添加 BFR 可以显著提高最差群体准确率(WGA),具体表现为最高提升了 26.0 个百分点。
- 主要贡献
- BiasFlow 工具和 BFR 方法能够在小规模基准测试中改善或保持平均 WGA,最高可提升 26.0 个百分点。
- 意义与局限
- 该工具和方法有助于提高模型的公平性和鲁棒性,但仅在小规模数据集上验证了其有效性,需要进一步在大规模数据集上进行测试。