CausalDynamics: A large-scale benchmark for structural discovery of dynamical causal models

NeurIPS 2025

A comprehensive benchmark framework designed to rigorously evaluate state-of-the-art causal discovery algorithms for dynamical systems.

Key Features

1️⃣ Large-Scale Benchmark. Systematically evaluate state-of-the-art causal discovery algorithms on thousands of graph challenges with increasing difficulty.

2️⃣ Customizable Data Generation. Scalable, user-friendly generation of increasingly complex coupled ordinary and stochastic systems of differential equations

3️⃣ Diverse Challenges. From simple chaotic systems to modular causal coupling of dynamical systems, including optional noise, confounding, time lags, and even climate model dynamics.

Abstract: Causal discovery for dynamical systems poses a major challenge in fields where active interventions are infeasible. Most methods used to investigate these systems and their associated benchmarks are tailored to deterministic, low-dimensional and weakly nonlinear time-series data. To address these limitations, we present CausalDynamics, a large-scale benchmark and extensible data generation framework to advance the structural discovery of dynamical causal models. Our benchmark consists of true causal graphs derived from thousands of coupled ordinary and stochastic differential equations as well as two idealized climate models. We perform a comprehensive evaluation of state-of-the-art causal discovery algorithms for graph reconstruction on systems with noisy, confounded, and lagged dynamics. CausalDynamics consists of a plug-and-play, build-your-own coupling workflow that enables the construction of a hierarchy of physical systems. We anticipate that our framework will facilitate the development of robust causal discovery algorithms that are broadly applicable across domains while addressing their unique challenges.

Datasets

You can generate your own dataset (see getting started), but you can also download our preprocessed ones directly from HuggingFace:

wget https://huggingface.co/datasets/kausable/CausalDynamics/resolve/main/process_causaldynamics.py
python process_causaldynamics.py

Installation

Using pip

CausalDynamics is available on PyPi, so you can use pip to install causaldynamics, which currently requires Python version 3.10.

pip install causaldynamics

If you use conda, please use the following commands:

conda create --name venv python=3.10
conda activate venv
pip install causaldynamics

Using pdm

Clone the repository and install it using pdm:

git clone https://github.com/kausable/CausalDynamics.git
cd CausalDynamics
pdm install

You can test whether the installation succeded by creating some coupled causal model data:

$(pdm venv activate)
python src/causaldynamics/creator.py --config config.yaml

You find the output at output/<timestamp> as default location.

Getting Started

• Challenge
• Quickstart
• CausalDynamics
• Benchmark
  • Simple Dynamics
  • Coupled Dynamics
  • Climate Dynamics
• Troubleshoot

Benchmarking

• Baseline
• Evaluation
• Leaderboard

Citation

If you find any of the code and dataset useful, feel free to acknowledge our work through:

@article{herdeanu2025causaldynamics,
  title={CausalDynamics: A large-scale benchmark for structural discovery of dynamical causal models},
  author={Herdeanu, Benjamin and Nathaniel, Juan and Roesch, Carla and Buch, Jatan and Ramien, Gregor and Haux, Johannes and Gentine, Pierre},
  journal={arXiv preprint arXiv:2505.16620},
  year={2025}
}