Dataset · Hugging Face

McGill-NLP/SpeechJBB

Name: SpeechJBB
Creator: McGill-NLP
License: other
Keywords: language:en, language:de, language:es, language:fr, language:it, license:other, size_categories:10K<n<100K, format:audiofolder, modality:audio, library:datasets, library:mlcroissant, arxiv:2606.06037, region:us, audio, speech

SpeechJBB Dataset Description SpeechJBB is an audio benchmark for evaluating safety alignment and comprehension in large audio language models (LALMs) under multilingual and code-switched speech. It is designed to test whether models consistently refuse harmful spoken requests when prompts are expressed in monolingual speech, code-switched speech, and code-switched speech with natural-sounding pseudo-word obfuscation. The benchmark extends JailbreakBench into… See the full description on the dataset page: https://huggingface.co/datasets/McGill-NLP/SpeechJBB.

View on Hugging FaceReport an issue

Author / organization	McGill-NLP
License (as declared)	other
Downloads	22.1K (as reported by Hugging Face)
Likes	4
Created	14 May 2026
Last updated at source	7 July 2026

audiospeechmultilingualcode-switchingsafetyred-teamingjailbreak-evaluationlarge-audio-language-modelssynthetic-speechbenchmark

Embed this entry

<iframe src="https://silytics.si/embed/item/b9ec7461-9307-4a3f-b797-19835a15e248" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Free, no tracking. More about feeds and embeds.

More from McGill-NLP

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 15 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.