dataset_info:
features:
- name: prompt
dtype: string - name: response
dtype: string - name: violation
dtype: string - name: category
dtype: string
splits: - name: test
num_bytes: 3187933
num_examples: 2873
download_size: 1755920
dataset_size: 3187933
configs: - config_name: default
data_files:- split: test
path: data/test-*
license: cc-by-nc-4.0
task_categories:
- split: test
- text-classification
language: - en
size_categories: - 1K<n<10K
tags: - safety
- content moderation
- LLM safety
- toxicity detection
Roblox Guard-Eval Dataset
We developed a custom high-quality evaluation dataset across Roblox’s content safety taxonomy—representing 25 subcategories. This evaluation set is created by internal red-teaming, where we test the system by simulating adversarial attacks to look for vulnerabilities, and doesn’t contain user-generated or personal data. This evaluation dataset contains prompt and response pairs with the responses hand-labeled by a set of policy experts to help ensure quality. It spans a wide spectrum of violation types, helping us create more precise and meaningful labels for evaluation. The final evaluation set includes 2,873 examples. This evaluation dataset, which features an extensible safety taxonomy to help benchmark LLM guardrails and moderation systems.
The LLM responses were generated by prompting Llama-3.2-3B-Instruct.
Citation
If you are using this dataset, please cite it as:
@article{nandwana2025taxonomy,
title={Taxonomy-Adaptive Moderation Model with Robust Guardrails for Large Language Models},
author={Nandwana, Mahesh Kumar and Lim, Youngwan and Liu, Joseph and Yang, Alex and Notibala, Varun and Khanna, Nishchaie},
journal={arXiv preprint arXiv:2512.05339},
year={2025}
}