发表机构
Hologen AI; Mathematics Research Centre Academy of Athens; FAU Erlangen Nuremberg; Imperial College London; University of Edinburgh（全息人工智能公司; 雅典科学院数学研究中心; 埃尔朗根 - 纽伦堡大学; 伦敦帝国理工学院; 爱丁堡大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
研究已训练模型所依赖因果与虚假特征的问题，提出归一化灵敏度比率（NSR）方法，在结构化转移情况下可事后诊断。在特定线性结构因果模型下能精确识别，刻画了失败情况及有限样本率，实验验证了该方法在多方面的有效性。
AI 中文摘要
给定一个已经训练好的模型，它在因果关系上依赖哪些特征而非虚假特征？现有方法需要访问训练过程，无法事后回答这个问题。我们引入了归一化灵敏度比率（NSR），这是一种在结构化转移情况下针对该问题的事后、模型无关的诊断方法。在多站点临床数据或多批次基因组学等场景中，环境主要在虚假特征的均值上有所不同，而因果机制和因果边际保持稳定。在此情况下，因果特征在各环境中会引起恒定的模型灵敏度，而虚假特征则跟踪变化。NSR将此形式化为每个环境灵敏度的变异系数平方。在具有$K\ge3$个非退化环境的线性结构因果模型（SCM）下，NSR实现了精确识别。我们全面刻画了失败情况，给出了实践者评估该情况是否成立的定量标准。有限样本率在原假设下为$O_p(n^{-1})$，在备择假设下为$O_p(n^{-1/2})$。实验证实了在合成数据上的所有理论预测，在五个模型家族中显示出一致的排名，并在不修改任何训练模型的情况下，在共享单车数据上恢复了八个因果特征中的六个。
英文摘要
Given a model that is already trained, which features does it rely on causally versus spuriously? Existing methods require access to the training procedure and cannot answer this post-hoc. We introduce the \textbf{Normalised Sensitivity Ratio~(NSR)}, a post-hoc, model-agnostic diagnostic for this question under a structured-shift regime: environments differ primarily in the mean of spurious features while the causal mechanism and causal marginals remain stable, as in multi-site clinical data or multi-batch genomics. Within this regime, causal features induce constant model sensitivity across environments while spurious features track shift. NSR formalises this as the squared coefficient of variation of per-environment sensitivity. Under a linear structural causal model (SCM) with $K\ge3$ non-degenerate environments, NSR achieves exact identification (Theorem~1). We fully characterise failure: weak shifts ($O(\varepsilon^4)$ collapse), degenerate geometry, and proxy attenuation ($O((1-α)^4)$), giving practitioners quantitative criteria for assessing whether the regime holds. Finite-sample rates are $O_p(n^{-1})$ under the null and $O_p(n^{-1/2})$ under the alternative. Experiments confirm all theoretical predictions on synthetic data (area under the ROC curve [AUROC] $= 1.000$ under conditions satisfying the regime), show consistent rankings across five model families (Kendall $τ\ge0.529$), and recover six of eight causal features on bike-sharing data (Precision@7 $= 0.75$) without modifying any trained model.
