Test Time Training for Supervised Causal Learning

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Deng, Zizhen, Zhang, Jiaru, Ding, Rui, Bojun, Huang, Wang, Jinzhuo, Fu, Qiang, Han, Shi, Zhang, Dongmei
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866910270215946240
author Deng, Zizhen
Zhang, Jiaru
Ding, Rui
Bojun, Huang
Wang, Jinzhuo
Fu, Qiang
Han, Shi
Zhang, Dongmei
author_facet Deng, Zizhen
Zhang, Jiaru
Ding, Rui
Bojun, Huang
Wang, Jinzhuo
Fu, Qiang
Han, Shi
Zhang, Dongmei
contents Supervised Causal Learning (SCL) has shown promise in causal discovery by framing it as a supervised learning problem. However, it suffers from significant out-of-distribution generalization challenges. We reveal three limitations of previous SCL practices: a significant performance gap between synthetic benchmarks and real-world data, fragility to distribution shifts, and failure in compositional generalization, collectively questioning its real-world applicability. To address this, we propose Test-Time Training for Supervised Causal Learning (TTT-SCL), a novel framework that dynamically generates training sets explicitly aligned with any specific test instance. We demonstrate the correlation between TTT-SCL and score-based methods, and design an efficient module for generating training sets based on the classic scoring function. Experiments on synthetic benchmarks, pseudo-real and real-world datasets demonstrate that TTT-SCL significantly outperforms existing SCL and traditional causal discovery methods.
format Preprint
id arxiv_https___arxiv_org_abs_2605_30015
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Test Time Training for Supervised Causal Learning
Deng, Zizhen
Zhang, Jiaru
Ding, Rui
Bojun, Huang
Wang, Jinzhuo
Fu, Qiang
Han, Shi
Zhang, Dongmei
Machine Learning
Artificial Intelligence
Supervised Causal Learning (SCL) has shown promise in causal discovery by framing it as a supervised learning problem. However, it suffers from significant out-of-distribution generalization challenges. We reveal three limitations of previous SCL practices: a significant performance gap between synthetic benchmarks and real-world data, fragility to distribution shifts, and failure in compositional generalization, collectively questioning its real-world applicability. To address this, we propose Test-Time Training for Supervised Causal Learning (TTT-SCL), a novel framework that dynamically generates training sets explicitly aligned with any specific test instance. We demonstrate the correlation between TTT-SCL and score-based methods, and design an efficient module for generating training sets based on the classic scoring function. Experiments on synthetic benchmarks, pseudo-real and real-world datasets demonstrate that TTT-SCL significantly outperforms existing SCL and traditional causal discovery methods.
title Test Time Training for Supervised Causal Learning
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2605.30015