Search papers, labs, and topics across Lattice.
The paper identifies limitations of Supervised Causal Learning (SCL) regarding out-of-distribution generalization, particularly on real-world data and under distribution shifts. To address this, they introduce Test-Time Training for SCL (TTT-SCL), which dynamically generates training sets tailored to each test instance. Experiments show TTT-SCL outperforms existing SCL and traditional causal discovery methods across various datasets.
SCL's promise falters in the real world, but dynamically adapting training data to each test instance can bridge the gap.
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.
