dataset_info:
features:
- name: instruction
dtype: string - name: response
dtype: string - name: prompt_label
dtype: string - name: response_label
dtype: string
splits: - name: train
num_bytes: 135329915
num_examples: 39949
download_size: 68654016
dataset_size: 135329915
configs: - config_name: default
data_files:- split: train
path: data/train-*
task_categories:
- split: train
- text-generation
- text-classification
language: - en
tags: - safety
- alignment
- reasoning
arxiv: 2601.23143
THINKSAFE Dataset
This dataset is part of the THINKSAFE: Self-Generated Safety Alignment for Reasoning Models project.
Dataset Description
This dataset contains safety-aligned training data for reasoning models, specifically generated using the Qwen3-0.6B model with WildGuard safety evaluation. It includes instructions, responses, and safety labels for both prompts and responses.
Dataset Structure
The dataset contains 39,949 examples with the following fields:
instruction: The input promptresponse: The model's generated responseprompt_label: Safety label for the promptresponse_label: Safety label for the response
Citation
If you use this dataset, please cite:
@article{lee2025thinksafe,
title={THINKSAFE: Self-Generated Safety Alignment for Reasoning Models},
author={Lee, Seanie and others},
journal={arXiv preprint arXiv:2601.23143},
year={2025}
}