Dataset · jackhhao
Jailbreak prompt classification
jackhhao/jailbreak-classification
LLM prompts labeled as jailbreak or benign to train harmful request detectors.
What's inside
A dataset for classifying LLM prompts as jailbreak or benign. Each example features a prompt field and a type label. Collected to help identify and prevent harmful jailbreak prompts during LLM interaction; source jailbreak prompts are listed in the card.
What you can do with it
- Training a jailbreak prompt classifier
- Filtering LLM inputs before inference
- Checking chatbot moderation
- Tasks
- Text classification
- Languages
- English
- Size
- 1K–10K rows
- Downloads this month
- 6.8K
- Likes
- 110
Models trained on it
- protectai/deberta-v3-base-prompt-injection-v2Text classification 380.7K
- patronus-studio/wolf-defender-prompt-injectionText classification 10K
- protectai/deberta-v3-small-prompt-injection-v2Text classification 6.9K
- patronus-studio/wolf-defender-prompt-injection-smallText classification 3.8K
- jackhhao/jailbreak-classifierText classification 1.8K
- codeintegrity-ai/promptguardText classification 1.4K
- mradermacher/Spellguard-1.0-8B-GGUF 934
- mradermacher/GPT-Distill-Qwen3-8B-Spellguard-SLERP-GGUF 479
How to load it
from datasets import load_dataset
ds = load_dataset("jackhhao/jailbreak-classification")