VoxSafeBench

VoxSafeBench Demopage: demopage/ Code: This dataset is uploaded as raw files (JSONL + audio), not parquet. Subset: Safety-tier1 Split: No jailbreak (8708 samples) Columns: system prompt, clean audio file name, diverse audio file name, transcript, super category, task type, language, query Split: Singleturn jailbreak (2516 samples) Columns:… See the full description on the dataset page:

Source: Hugging Face Hub (YuxiangW/VoxSafeBench). Metadata imported from the dataset’s Hub tags.