Dataset · Hugging Face

allenai/wildjailbreak

Name: wildjailbreak
Creator: allenai
License: odc-by
Keywords: task_categories:text-generation, language:en, license:odc-by, size_categories:1K<n<10K, format:csv, modality:image, modality:text, library:datasets, library:pandas, library:mlcroissant, library:polars, arxiv:2406.18510, region:us, ai safety, jailbreak

WildJailbreak Dataset Card WildJailbreak is an open-source synthetic safety-training dataset with 262K vanilla (direct harmful requests) and adversarial (complex adversarial jailbreaks) prompt-response pairs. In order to mitigate exaggerated safety behaviors, WildJailbreaks provides two contrastive types of queries: 1) harmful queries (both vanilla and adversarial) and 2) benign queries that resemble harmful queries in form but contain no harmful intent. Vanilla Harmful: direct… See the full description on the dataset page: https://huggingface.co/datasets/allenai/wildjailbreak.

View on Hugging FaceReport an issue

Author / organization	allenai
License (as declared)	odc-by
Downloads	6.8K (as reported by Hugging Face)
Likes	163
Created	18 June 2024
Last updated at source	8 August 2024

ai safetyjailbreaksafety trainingred-teaming

Embed this entry

<iframe src="https://silytics.si/embed/item/0c460f26-cff0-4177-b80e-be8299066c9c" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Free, no tracking. More about feeds and embeds.

Popularity, last 2 days

6,778 now (-103 since 2026-10-16). Counts as reported by the source, snapshotted daily by Silytics.

Connected in the knowledge graph

vllm-sr/Decision-1.0-Route-0.6BModel · used to train

Explore this entry in the graph

More from allenai

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 17 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.