ProximalFM: Amortized Proximal Causal Inference under Hidden Confounding

📄 arXiv: 2610.08078v1 📥 PDF

作者: Christophe Muller, Ayub Kharel, Alex Luedtke, Chan Park, Eric Tchetgen Tchetgen, Juan L. Gamella, Rahul Krishnan, Ricardo Silva, Jakob Zeitler

分类: stat.ML, cs.LG

发布日期: 2026-10-06

💡 一句话要点

提出ProximalFM以解决隐藏混杂因素下的因果推断问题

🎯 匹配领域: 支柱九：具身大模型 (Embodied Foundation Models)

关键词: 因果推断 贝叶斯推断 隐藏混杂 代理变量 条件平均处理效应 机器学习 结构因果模型

📋 核心要点

  1. 现有因果推断方法在面对未测量混杂因素时，常常无法有效识别因果效应，导致结果不可靠。
  2. ProximalFM模型通过使用代理变量和贝叶斯推断，简化了因果效应的估计过程，能够有效处理隐藏混杂问题。
  3. 在多种Proximal场景下，ProximalFM模型在CATE估计性能上表现优异，尤其在潜在混杂因素显著且代理信息较弱时，优势明显。

📝 摘要（中文）

标准因果识别方法通常假设没有未测量的混杂因素，而在相关混杂因素未被观察时可能失效。Proximal因果推断利用代理变量来识别隐藏混杂下的效应。然而，非参数的Proximal估计在实践中面临挑战，恢复条件平均处理效应（CATE）需要解决一个数据需求高、超参数敏感且优化不稳定的病态积分方程。本文提出的ProximalFM模型，通过在合规结构因果模型上训练合成数据，简化了任务，将贝叶斯操作符反演的复杂性转化为单次变换器前向传播，显著提高了CATE的估计性能。

🔬 方法详解

问题定义：本文旨在解决在隐藏混杂因素影响下的因果推断问题。现有方法在未测量混杂因素存在时，往往无法准确识别因果效应，导致推断结果的可靠性下降。

核心思路：ProximalFM模型利用代理变量和贝叶斯推断来简化因果效应的估计过程。通过在合成数据上训练，模型能够有效地处理潜在的混杂因素，并将复杂的贝叶斯操作符反演转化为简单的前向传播。

技术框架：该模型的整体架构包括数据生成、模型训练和推断三个主要阶段。首先，生成合成数据以模拟合规结构因果模型；其次，使用前向传播的方式进行模型训练；最后，通过单次前向传播实现快速推断。

关键创新：ProximalFM的主要创新在于其将贝叶斯后验分布的估计与传统的点估计方法区分开来，特别是针对CATE的后验分布进行建模，显著提高了估计的准确性和稳定性。

关键设计：模型设计中采用了Monte Carlo方法来估计oracle CATEs，并引入了新的PFN变体以处理额外的随机误差。模型在超参数设置和损失函数设计上进行了优化，以确保在不同数据集上的一致性表现。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

在多种Proximal场景下，ProximalFM模型在CATE估计性能上表现出色，相较于传统方法，其在潜在混杂因素显著且代理信息较弱的情况下，CATE估计的准确性提升了显著，且无需针对特定数据集进行调优，推断速度也得到了显著提高。

🎯 应用场景

ProximalFM模型在医疗、社会科学和经济学等领域具有广泛的应用潜力，尤其是在需要处理未测量混杂因素的因果推断任务中。其快速推断能力和高准确性使其在实际决策支持系统中具有重要价值，未来可能推动因果推断技术的进一步发展。

📄 摘要（原文）

Standard causal identification methods often assume no unmeasured confounding and can fail when relevant confounders are unobserved. Proximal causal inference instead uses proxy variables to identify effects under hidden confounding. However, nonparametric proximal estimation can be challenging in practice: recovering causal estimands such as the conditional average treatment effect (CATE) requires solving an ill-posed integral equation that is data-hungry, hyperparameter-sensitive, and optimization-unstable. Bayesian inference for such models provides a desirable alternative, mitigating these difficulties by regularizing through the prior. However, computing a posterior is itself challenging, as a typical likelihood function will include latent variables. Following the recent success of tabular foundation models in backdoor, instrumental variable, and frontdoor settings, we propose that prior-data fitted networks (PFNs) are uniquely suited to resolve this bottleneck. Indeed, by training on synthetic data sampled from compliant structural causal models with access to oracle counterfactuals, we simplify the task substantially, amortizing the implied Bayesian operator inversion into a single transformer forward pass. Compared to prior literature that focuses primarily on point estimation, our model, ProximalFM, explicitly targets the Bayesian posterior distribution of the CATE. One unique aspect of this problem is that we need to provide Monte Carlo estimates of the oracle CATEs, leading to a novel variation of PFNs that accounts for the added stochastic error. Across a diverse suite of proximal regimes, ProximalFM achieves consistently strong CATE-estimation performance without dataset-specific tuning, with its largest advantage when latent confounding is substantial and the proxies are weakly informative; it also provides fast inference through a single amortized forward pass.