![[2602.19131] Test-Time Learning of Causal Structure from Interventional Data](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.19131] Test-Time Learning of Causal Structure from Interventional Data
Summary
The paper presents TICL, a novel method for causal structure learning from interventional data, enhancing generalization across diverse settings through self-augmentation and joint causal inference.
Why It Matters
Causal discovery is crucial in various fields, including healthcare and social sciences. TICL addresses the challenge of generalization in supervised causal learning, potentially improving decision-making processes in uncertain environments.
Key Takeaways
- TICL combines Test-Time Training with Joint Causal Inference for improved causal structure learning.
- The method generates instance-specific training data at test time to mitigate distribution shifts.
- Experiments show TICL's effectiveness in causal discovery and intervention target detection.
Computer Science > Machine Learning arXiv:2602.19131 (cs) [Submitted on 22 Feb 2026] Title:Test-Time Learning of Causal Structure from Interventional Data Authors:Wei Chen, Rui Ding, Bojun Huang, Yang Zhang, Qiang Fu, Yuxuan Liang, Han Shi, Dongmei Zhang View a PDF of the paper titled Test-Time Learning of Causal Structure from Interventional Data, by Wei Chen and 7 other authors View PDF HTML (experimental) Abstract:Supervised causal learning has shown promise in causal discovery, yet it often struggles with generalization across diverse interventional settings, particularly when intervention targets are unknown. To address this, we propose TICL (Test-time Interventional Causal Learning), a novel method that synergizes Test-Time Training with Joint Causal Inference. Specifically, we design a self-augmentation strategy to generate instance-specific training data at test time, effectively avoiding distribution shifts. Furthermore, by integrating joint causal inference, we developed a PC-inspired two-phase supervised learning scheme, which effectively leverages self-augmented training data while ensuring theoretical identifiability. Extensive experiments on bnlearn benchmarks demonstrate TICL's superiority in multiple aspects of causal discovery and intervention target detection. Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI) Cite as: arXiv:2602.19131 [cs.LG] (or arXiv:2602.19131v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2602.19131 F...
