不修改训练好的模型，就能分辨出哪些特征是真正重要的因果特征。
From Training to Deployment: Post-Hoc Causal Feature Identification via Sensitivity Ratios
- 通过敏感度变异率（NSR）检测各环境下的模型响应变化，区分因果与虚假特征。
- 在满足条件的多中心数据中，能准确识别出87.5%的因果特征（精度@7=0.75）。
- 适用于临床、基因组等多环境数据，无需重训练模型，适合实际部署场景。
针对已训练完成的模型，如何判断其依赖的是因果特征还是虚假相关特征？现有方法需访问训练过程，无法实现事后诊断。本文提出一种名为归一化敏感度比率（NSR）的后验、模型无关诊断方法，适用于结构化变化场景：即虚假特征的均值在不同环境间变化，而因果机制和因果边际保持稳定，如多中心临床数据或分批基因组数据。在此设定下，因果特征导致模型敏感度在各环境中保持恒定，而虚假特征随环境变化。NSR将这一差异形式化为每环境敏感度的平方变异系数。在具有K≥3个非退化环境的线性结构因果模型中，NSR可实现精确识别（定理1）。我们全面刻画了失败情形：弱扰动（O(ε⁴)坍缩）、退化几何结构及代理衰减（O((1−α)⁴)），为实践者提供判断该设定是否成立的量化标准。有限样本下，零假设下收敛速率为O_p(n⁻¹)，备择假设下为O_p(n⁻¹/²)。实验验证了所有理论预测：合成数据上受试者工作特征曲线下面积（AUROC）达1.000；五种模型家族间排名一致性高（Kendall τ≥0.529）；在共享单车数据中成功恢复8个因果特征中的6个（精度@7=0.75），且未修改任何预训练模型。
原文摘要 · Abstract (English)
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.
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。