Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
UnCLe: Towards Scalable Dynamic Causal Discovery in Non-linear Temporal Systems
Authors: Tingzhu Bi, Yicheng Pan, Xinrui Jiang, Huize Sun, Meng Ma, Ping Wang
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments demonstrate that Un CLe not only outperforms state-of-the-art baselines on static causal discovery benchmarks but, more importantly, exhibits a unique capability to accurately capture and represent evolving temporal causality in both synthetic and real-world dynamic systems (e.g., human motion). |
| Researcher Affiliation | Academia | Tingzhu Bi, Yicheng Pan, Xinrui Jiang, Huize Sun, Meng Ma , Ping Wang Peking University EMAIL, EMAIL EMAIL, EMAIL |
| Pseudocode | Yes | The detailed algorithm for dynamic causal discovery via temporal perturbation is listed as Algorithm 1 in Appendix L. |
| Open Source Code | Yes | Code & Datasets: https://github.com/etigerstudio/uncle-causal-discovery |
| Open Datasets | Yes | Code & Datasets: https://github.com/etigerstudio/uncle-causal-discovery |
| Dataset Splits | No | The paper discusses evaluation across 'replicas' of datasets and 'segments' for dynamic causality but does not specify explicit training, validation, and test splits (e.g., percentages or sample counts) for the time series data used to train the models or evaluate the causal discovery. It mentions evaluating inferred structures against true structures. |
| Hardware Specification | No | The paper's text does not contain specific details regarding the hardware (e.g., GPU models, CPU types, or memory) used for running the experiments. While the NeurIPS checklist indicates that hardware configuration is provided, these details are absent from the body of the paper. |
| Software Dependencies | No | The paper mentions several software libraries and frameworks used by baseline methods (e.g., 'statsmodels library', 'tigramite', 'lingam', 'causalnex', 'CUTS+', 'JRNGC') but does not specify their version numbers or the version of the underlying programming language (e.g., Python) or deep learning frameworks used for Un CLe. |
| Experiment Setup | Yes | Table 9: Hyperparameter Settings for Un CLe. Dataset Lag Kernel Size TCN Blocks Kernel Filters Recon. Epochs Joint Epochs Learning Rate Lorenz#1 1 8 6 20 1,000 2,000 5e-3 Lorenz#2 1 6 8 12 1,000 2,500 2e-3 Lorenz#3 1 3 6 18 500 2,500 1e-3 f MRI 1 6 8 12 1,000 2,000 1e-5 NC8 1 8 6 20 1,000 2,000 3e-4 FINANCE 2 2 3 24 500 10,000 3e-4 ND8 1 8 6 20 1,000 2,000 3e-4 TVSEM 1 3 4 8 500 2,500 2e-3 |