Saved in:
Bibliographic Details
Main Authors: Bing, Simon, Wahl, Jonas, Runge, Jakob
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2603.08682
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912968187314176
author Bing, Simon
Wahl, Jonas
Runge, Jakob
author_facet Bing, Simon
Wahl, Jonas
Runge, Jakob
contents We introduce structural causal bottleneck models (SCBMs), a novel class of structural causal models. At the core of SCBMs lies the assumption that causal effects between high-dimensional variables only depend on low-dimensional summary statistics, or bottlenecks, of the causes. SCBMs provide a flexible framework for task-specific dimension reduction while being estimable via standard, simple learning algorithms in practice. We analyse identifiability in SCBMs, connect them to information bottlenecks in the sense of Tishby & Zaslavsky (2015), and illustrate how to estimate them experimentally. We also demonstrate the benefit of bottlenecks for effect estimation in low-sample transfer learning settings. We argue that SCBMs provide an alternative to existing causal dimension reduction frameworks like causal representation learning or causal abstraction learning.
format Preprint
id arxiv_https___arxiv_org_abs_2603_08682
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Structural Causal Bottleneck Models
Bing, Simon
Wahl, Jonas
Runge, Jakob
Machine Learning
We introduce structural causal bottleneck models (SCBMs), a novel class of structural causal models. At the core of SCBMs lies the assumption that causal effects between high-dimensional variables only depend on low-dimensional summary statistics, or bottlenecks, of the causes. SCBMs provide a flexible framework for task-specific dimension reduction while being estimable via standard, simple learning algorithms in practice. We analyse identifiability in SCBMs, connect them to information bottlenecks in the sense of Tishby & Zaslavsky (2015), and illustrate how to estimate them experimentally. We also demonstrate the benefit of bottlenecks for effect estimation in low-sample transfer learning settings. We argue that SCBMs provide an alternative to existing causal dimension reduction frameworks like causal representation learning or causal abstraction learning.
title Structural Causal Bottleneck Models
topic Machine Learning
url https://arxiv.org/abs/2603.08682