Large Causal Models for Temporal Causal Discovery

2026-02-20

AI Summary

This work addresses the limited generalizability of traditional causal discovery methods, which are typically constrained to modeling a single dataset and often underperform in high-dimensional temporal settings. To overcome these limitations, the authors propose the Large Causal Model (LCM) framework—the first to adapt the foundation model paradigm to temporal causal discovery. LCM leverages pretraining on a diverse mixture of synthetic and real-world time-series data, enabling it to handle high-dimensional variables, deep architectures, and efficient single-pass inference. Extensive evaluations demonstrate that LCM matches or surpasses both classical and neural baseline methods across synthetic, semi-synthetic, and real-world benchmarks, exhibiting particularly strong out-of-distribution generalization and computational efficiency.

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality
• Cognitive Modeling & Cognitive Systems: Other Foundations of Cognitive Modeling & Systems

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Causal discovery for both cross-sectional and temporal data has traditionally followed a dataset-specific paradigm, where a new model is fitted for each individual dataset. Such an approach limits the potential of multi-dataset pretraining. The concept of large causal models (LCMs) envisions a class of pre-trained neural architectures specifically designed for temporal causal discovery. Prior approaches are constrained to small variable counts, degrade with larger inputs, and rely heavily on synthetic data, limiting generalization. We propose a principled framework for LCMs, combining diverse synthetic generators with realistic time-series datasets, allowing learning at scale. Extensive experiments on synthetic, semi-synthetic and realistic benchmarks show that LCMs scale effectively to higher variable counts and deeper architectures while maintaining strong performance. Trained models achieve competitive or superior accuracy compared to classical and neural baselines, particularly in out-of-distribution settings, while enabling fast, single-pass inference. Results demonstrate LCMs as a promising foundation-model paradigm for temporal causal discovery. Experiments and model weights are available at https://github.com/kougioulis/LCM-paper/.

Problem

• temporal causal discovery
• large causal models
• multi-dataset pretraining
• scalability
• generalization

Innovation

• Large Causal Models
• Temporal Causal Discovery
• Pretrained Neural Architecture
• Out-of-Distribution Generalization
• Scalable Causal Inference