README.md

---
pretty_name: J
dataset_info:
- config_name: Github_medium
  features:
  - name: json_schema
    dtype: string
  - name: unique_id
    dtype: string
  splits:
  - name: train
    num_examples: 1
  - name: val
    num_examples: 1
  - name: test
    num_examples: 100
configs:
- config_name: Github_medium
  data_files:
  - split: train
    path: Github_medium/train-*
  - split: val
    path: Github_medium/val-*
  - split: test
    path: Github_medium/test-*
license: mit
task_categories:
- text-generation
---

This is a pruned eval dataset from [epfl-dlab/JSONSchemaBench](https://huggingface.co/datasets/epfl-dlab/JSONSchemaBench) for personal debugging purposes.

Below is the original model card.

***
# JSONSchemaBench

[![Paper](https://img.shields.io/badge/Paper-arXiv-blue)](https://arxiv.org/abs/2501.10868)
[![GitHub](https://img.shields.io/badge/Code-GitHub-blue)](https://github.com/guidance-ai/jsonschemabench)

JSONSchemaBench is a benchmark of **real-world JSON schemas** designed to evaluate **structured output generation** for Large Language Models (LLMs). It contains approximately **10,000 JSON schemas**, capturing diverse constraints and complexities.


```python
import datasets
from datasets import load_dataset

def main():
    # Inspect the available subsets of the dataset
    all_subsets = datasets.get_dataset_config_names("epfl-dlab/JSONSchemaBench")
    print("Available subsets:", all_subsets)
    # Example output: ['Github_easy', 'Github_hard', 'Github_medium', 'Github_trivial', 'Github_ultra', 'Glaiveai2K', 'JsonSchemaStore', 'Kubernetes', 'Snowplow', 'WashingtonPost', 'default']

    # Access a specific subset of the dataset
    subset_name = "Github_easy"
    github_easy = load_dataset("epfl-dlab/JSONSchemaBench", subset_name)
    print(f"Loaded subset '{subset_name}':", github_easy)

    # Load the entire dataset as a whole
    entire_dataset = load_dataset("epfl-dlab/JSONSchemaBench", "default")
    print("Loaded entire dataset:", entire_dataset)

if __name__ == "__main__":
    main()
```

## Update (March 31st, 2025)

To improve inference efficiency and streamline data collation, we’ve decided to drop a small number of exceptionally long samples from the dataset.

We’re using the `meta-llama/Llama-3.2-1B-instruct` tokenizer, and the filtering criteria are as follows:
- Github_easy: Samples longer than 1024 tokens — 5 out of 582 removed
- Github_medium: Samples longer than 2048 tokens — 7 out of 593 removed
- Github_hard: Samples longer than 8192 tokens — 4 out of 372 removed
- Other subsets are not touched

Since the number of discarded samples is minimal, this change is expected to have at most a 1% impact on results.


## ⚠️ Important Update (March 10th, 2025)

We have restructured the dataset to include train/val/test splits. If you downloaded the dataset before this date, you might encounter errors like `KeyError: 'Github_easy'`.

To fix this issue, please follow one of the options below:

1. Update How Subsets Are Accessed:
If you previously used:

```python
from datasets import load_dataset, concatenate_datasets, DatasetDict, Dataset

subset: DatasetDict = load_dataset("epfl-dlab/JSONSchemaBench")
subset["Github_easy"]
```
You can update it to:

```python
from datasets import load_dataset, concatenate_datasets, DatasetDict, Dataset

subset: DatasetDict = load_dataset("epfl-dlab/JSONSchemaBench", name="Github_easy")
subset: Dataset = concatenate_datasets([subset["train"], subset["val"], subset["test"]])
```

2. Load the Dataset in the Old Structure:
If you need the previous structure, you can use a specific revision:

```python
dataset = load_dataset("epfl-dlab/JSONSchemaBench", revision="e2ee5fdba65657c60d3a24b321172eb7141f8d73")
```

We apologize for the inconvenience and appreciate your understanding! 😊

## 📌 Dataset Overview
- **Purpose:** Evaluate the **efficiency** and **coverage** of structured output generation.
- **Sources:** GitHub, Kubernetes, API specifications, curated collections.
- **Schemas:** Categorized based on complexity and domain.

### 📊 Dataset Breakdown
| Dataset         | Category            | Count |
| --------------- | ------------------- | ----- |
| GlaiveAI-2K     | Function Call       | 1707  |
| Github-Trivial  | Misc                | 444   |
| Github-Easy     | Misc                | 1943  |
| Snowplow        | Operational API     | 403   |
| Github-Medium   | Misc                | 1976  |
| Kubernetes      | Kubernetes API      | 1064  |
| Washington Post | Resource Access API | 125   |
| Github-Hard     | Misc                | 1240  |
| JSONSchemaStore | Misc                | 492   |
| Github-Ultra    | Misc                | 164   |
| **Total**       |                     | 9558  |

## 📥 Loading the Dataset

```python
from datasets import load_dataset

dataset = load_dataset("epfl-dlab/JSONSchemaBench")
print(dataset)
```

## 🔍 Data Structure
Each dataset split contains:
- `"json_schema"`: The schema definition.
- `"unique_id"`: A unique identifier for the schema.


🚀 **For more details, check out the [paper](https://arxiv.org/abs/2501.10868).**

## 📚 Citation
```bibtex
@misc{geng2025jsonschemabench,
      title={Generating Structured Outputs from Language Models: Benchmark and Studies},
      author={Saibo Geng et al.},
      year={2025},
      eprint={2501.10868},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2501.10868}
}
```


## License

This dataset is provided under the [MIT License](https://opensource.org/licenses/MIT). Please ensure that you comply with the license terms when using or distributing this dataset.

## Acknowledgements

We would like to thank the contributors and maintainers of the JSON schema projects and the open-source community for their invaluable work and support.