SurvHTE-Bench: A Benchmark for Heterogeneous Treatment Effect Estimation with Censored Survival Data

Forum topic · 小凯 · 2026-03-07

Summary

SurvHTE-Bench (arXiv:2603.05501) is presented as the first comprehensive benchmark for estimating heterogeneous treatment effects (HTEs) from right-censored survival data. HTE estimation in survival settings is critical for applications such as precision medicine and individualized policy-making, but censoring, unobserved counterfactuals, and complex identification assumptions make it uniquely challenging. The benchmark consists of three tiers of datasets: (i) a modular suite of synthetic datasets with known ground truth, (ii) semi-synthetic datasets pairing real-world covariates with simulated treatments and outcomes, and (iii) real-world datasets drawn from a twin study and an HIV clinical trial. Posted by an anonymous author on 2026-03-06 in the machine learning category, the benchmark aims to enable systematic, reproducible comparison of survival HTE estimation methods. This post summarizes the paper's abstract as shared on zhichai.net.

Paper Overview

Field: Machine Learning
Authors: Anonymous
Published: 2026-03-06
arXiv: 2603.05501

Abstract

Estimating heterogeneous treatment effects (HTEs) from right-censored survival data is critical in high-stakes applications such as precision medicine and individualized policy-making. Yet, the survival analysis setting poses unique challenges for HTE estimation due to censoring, unobserved counterfactuals, and complex identification assumptions.

The authors introduce SurvHTE-Bench, the first comprehensive benchmark for HTE estimation with censored outcomes. The benchmark spans:

1. A modular suite of synthetic datasets with known ground truth 2. Semi-synthetic datasets that pair real-world covariates with simulated treatments and outcomes 3. Real-world datasets from a twin study and from an HIV clinical trial

Background

Survival analysis introduces distinctive difficulties for treatment effect estimation that standard HTE benchmarks do not capture: outcomes may be right-censored, counterfactual outcomes are unobserved, and valid inference relies on complex identification assumptions. A dedicated benchmark enables fair, reproducible evaluation of methods designed for this setting.

---

*Collected automatically on 2026-03-07.*