Residual-Based Causal Discovery via Pretrained In-Context Quantile Regression
Abstract
Residual-based causal discovery assumes that each variable is generated from its parents and an independent noise source. It estimates this noise and tests its independence from candidate causes. For identifiable functions such as additive-noise models (ANMs) and the broader class of location-scale noise models (LSNMs), this principle identifies the causal direction, as only the causal model admits this independence. Existing LSNM methods rely on parametric assumptions to obtain standardized residuals, risking model misspecification. In contrast, the RESIT-framework allows for nonparametric estimation of residuals to iteratively identify sink nodes, but was proposed for ANMs. To close this gap, we propose Loscar, an extension of RESIT to LSNMs that estimates the conditional mean and scale via quantile regression, enabling pretrained tabular foundation models as plug-in estimators. We compare it with a feasible generalized least squares (FGLS) variant, which is consistent given suitable basis functions. Across bivariate and multivariate causal-discovery benchmarks, in-context quantile regression substantially outperforms FGLS with flexible function classes, particularly for multivariate causes.
Related
- Quantile-Only Regression: Inference and Partial Identification under Quantile Disclosure
- Predictive-Context LOO: Bayesian Model Comparison with Pretrained Predictive Learners
- Clouded Representatives and Score Matching for Distributed Quantile Regression
- A New Framework for Estimation of Unconditional Quantile Treatment Effects: The Residualized Quantile Regression (RQR) Model
- Conformalized Quantile Regression and Minimax Limits of Fixed-Score Calibration under Known Covariate Shift
BibTeX
@misc{klippert2026residualbased,
title = {{Residual-Based Causal Discovery via Pretrained In-Context Quantile Regression}},
author = {Daniel Klippert and Quang Nhat Huynh and Alexander Marx},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.29530},
eprint = {2610.29530},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.29530},
}
RIS
TY - UNPB
TI - Residual-Based Causal Discovery via Pretrained In-Context Quantile Regression
AU - Daniel Klippert
AU - Quang Nhat Huynh
AU - Alexander Marx
PY - 2026
DA - 2026/10/21
PB - arXiv
DO - 10.48550/arxiv.2610.29530
UR - https://arxiv.org/abs/2610.29530
AB - Residual-based causal discovery assumes that each variable is generated from its parents and an independent noise source. It estimates this noise and tests its independence from candidate causes. For identifiable functions such as additive-noise models (ANMs) and the broader class of location-scale noise models (LSNMs), this principle identifies the causal direction, as only the causal model admits this independence. Existing LSNM methods rely on parametric assumptions to obtain standardized residuals, risking model misspecification. In contrast, the RESIT-framework allows for nonparametric estimation of residuals to iteratively identify sink nodes, but was proposed for ANMs. To close this gap, we propose Loscar, an extension of RESIT to LSNMs that estimates the conditional mean and scale via quantile regression, enabling pretrained tabular foundation models as plug-in estimators. We compare it with a feasible generalized least squares (FGLS) variant, which is consistent given suitable basis functions. Across bivariate and multivariate causal-discovery benchmarks, in-context quantile regression substantially outperforms FGLS with flexible function classes, particularly for multivariate causes.
ER -