| --- |
| pretty_name: CausalMix v1 |
| configs: |
| - config_name: small |
| data_files: |
| - split: test |
| path: "graphs/small/*.csv" |
| - config_name: large |
| data_files: |
| - split: test |
| path: "graphs/large/*.csv" |
| - config_name: synthetic |
| data_files: |
| - split: test |
| path: "graphs/synthetic/*.csv" |
| --- |
| |
| # Anonymous Data |
|
|
| This dataset contains benchmark files for evaluating language models on causal-discovery tasks. |
|
|
| The `small` and `synthetic` configs are available through the Hugging Face dataset viewer. The `large` config is provided as direct CSV files under `graphs/large/`; these files are large enough that automatic viewer/Croissant conversion may omit them, so the NeurIPS Croissant metadata is completed manually from the hosted CSV file layout. |
|
|
