license: cc-by-nc-sa-4.0
task_categories:
- text-classification
language:
- en
tags:
- safety
- alignment
- content-moderation
- guardrail
- supervision
pretty_name: BELLS-O Content Moderation Output Dataset
size_categories:
- 1K<n<10K
language_creators:
- machine-generated
source_datasets:
- extended|bells-o-project/content-moderation-input
multilinguality:
- monolingual
extra_gated_prompt: You agree not to use the dataset or information included in it to cause harm.
extra_gated_fields:
Name of the Company or Institution you are affiliated with, or Independent: text
Institutional/Organizational email address:
type: text
optional: true
Country: country
Describe your intended use in 2-3 sentences (insufficient detail = declined): text
I agree to use this dataset for non-commercial use ONLY: checkbox
I read the 'Access Guidelines' section in the README: checkbox
Access Guidelines - READ THIS BEFORE REQUESTING ACCESS!
Access is only granted to identifiable individuals with proper reason to use this sensitive data.
If any other dataset could be used to accomplish your goal, this does not count as a proper reason. Half sentences and bullet points do not suffice and will be declined. Proper reasons include anything that showcases your specific need for this exact dataset.
Content Moderation Output Dataset
Overview
This dataset contains 1,300 model responses corresponding to 1,300 prompts from the BELLS-Operational content moderation input benchmark.(Excluding Hate Speech) Each response has undergone a two-stage generation pipeline: initial generation followed by paraphrasing to remove model fingerprints, ensuring robust evaluation of supervision systems.
Dataset Statistics
| Category | Responses |
|---|---|
| Benign | 300 |
| CBRN | 100 |
| Cyber | 100 |
| Harm to Minors | 100 |
| Harmful Manipulation | 100 |
| Illegal Activities | 100 |
| Integrity & Quality | 100 |
| Physical Harm | 100 |
| Privacy | 100 |
| Self-Harm | 100 |
| Sexual Content | 100 |
| TOTAL | 1,300 |
Taxonomy
Our taxonomy covers 11 harmful categories aligned with EU AI Act systemic risk categories and GPAI responsible use policies and 1 benign category to measure the harmful refusal rate:
Harmful Categories (11)
- CBRN: Chemical, biological, radiological, nuclear weapons and dangerous materials
- Cyber: Malware, hacking, exploits, and cybersecurity threats
- Harm to Minors: Content endangering, exploiting, or sexualizing individuals under 18
- Harmful Manipulation: Scams, fraud, phishing, psychological manipulation, misinformation
- Illegal Activities: Drug trafficking, weapons trafficking, financial crimes, human trafficking
- Integrity & Quality: Academic dishonesty, plagiarism, credential falsification, spam
- Physical Harm: Violence instructions, weapons, explosives, dangerous activities
- Privacy: Doxxing, unauthorized surveillance, personal information disclosure, stalking
- Self-Harm: Suicide, eating disorders, self-injury promotion or instructions
- Sexual Content: Explicit material, pornography, non-consensual intimate content
Benign Category (1)
Benign: Legitimate responses to acceptable requests covering cooking, education, travel, technology, creative writing, and general knowledge.
Methodology
Two-Stage Generation Pipeline
Stage 1: Initial Generation
Model generates responses to all 1,400 input prompts from the dataset using Dolphin Mistral 24B Venice Edition (dphn/Dolphin-Mistral-24B-Venice-Edition). This captures model behavior across all harmful categories and benign content.
Stage 2: Paraphrasing
All generated responses undergo paraphrasing using Dolphin 2.2 70B (dphn/dolphin-2.2-70b) to:
- Remove model fingerprints: Eliminate distinctive stylistic patterns that could allow some supervision systems to identify the harmful samples more easily
- Preserve semantic content: Maintain the actual information and intent of the response
- Enable fair evaluation: Ensure supervisors assess what was said, not who said it
Quality Control
- Semantic preservation verified through automated checks
- Sample validation by human reviewers
- Comparison between original and paraphrased responses enables quality assessment
Models Used
Generation Model:
- Dolphin Mistral 24B Venice Edition. Hugging Face.
Paraphrasing Model:
- Hartford, E., & Cognitive Computations (2023). Dolphin 2.2 70B. Hugging Face
Purpose
Evaluate supervision systems on content moderation responses under operational constraints.
This output dataset enables assessment of how supervision systems classify model-generated responses across harmful and benign categories. The paraphrasing stage ensures evaluation focuses on content rather than model-specific writing patterns.
Relationship to Input Dataset
This output dataset is the companion response dataset to the input prompt dataset, which contains:
- 1,400 evaluation prompts
- 12-category taxonomy
- Three data sources (380 AI-generated, 620 extracted, 400 handcrafted)
Together, these datasets enable end-to-end evaluation:
- Input dataset provides standardized test prompts
- Output dataset provides model responses with fingerprint removal
- Supervision systems are evaluated on detecting harmful content in responses
Contact
- Leonhard Waibl: leonhard.waibl{at}student[dot]tugraz.at
- Felix Michalak: felix[at]michalax{.}de
- Hadrien Mariaccia: hadrien{at}securite-ia[dot]fr
Citation
If you use this dataset in your research, please cite:
@dataset{bells_content_moderation_output_dataset_2026,
title={BELLS-O: Content Moderation Output Dataset},
author={Waibl, Leonhard and Michalak, Felix and Mariaccia, Hadrien},
year={2026},
publisher={Hugging Face},
howpublished={\url{https://huggingface.co/datasets/centrepourlasecuriteia/content-moderation-output-dataset}}
}
Part of BELLS-Operational • SPAR Fall 2025 • CeSIA
Companion dataset to the Content Moderation Input Dataset https://huggingface.co/datasets/bells-o-project/content-moderation-input