Causal Chamber™: Light Tunnel Interventions v1

These datasets were created in the Causal Chamber “Light Tunnel” (Gamella et al., 2025). The Light Tunnel is an apparatus to sample real data under controlled physical conditions. For more information, see https://causalchamber.ai/. The dataset collection contains experiments in which single-target interventions of different strengths are performed on all the light tunnel's manipulable variables. The light tunnel is operating in its standard configuration. Each experiment corresponds to a different intervention, where the target distribution is shifted with respect to its reference distribution, given by the dataset uniform_reference.csv. For a detailed description of the data and each experiment, see https://github.com/juangamella/causal-chamber/tree/main/datasets/lt_interventions_standard_v1 Task: The dataset collection can be used to study causal discovery and causal inference methods. Summary: Size of collection: 59 Datasets with 46 columns of size 1000 (10^4 for uniform_reference.csv) Task: Causal Discovery Problem / Causal Inference Problem Data Type: Mixed Data Dataset Scope: Collection of Datasets Ground Truth: Known Graph Temporal Structure: Static Data License: CC BY 4.0 Missing Values: No Missing Values Missingness Statement: There are no missing values. Collection: The file collection.html contains an overview of all interventions generating the datasets. Features: The file variables.csv contains a brief description of each variable (column) in the .csv files; see appendix II of the manuscript.pdf for more details. Files: lt_interventions_standard_v1.zip: Collection of the interventional and the reference Dataset collection.html: An overview of all interventions generating the datasets. variables.csv: A brief description of each variable. arxiv_gamella_2025.pdf: ArXiv version of Gamella et al., 2025, for reference. ground_truths.pdf: See Figure a for the ground truth DAG. License: lt_interventions_standard_v1.zip: CC BY 4.0