datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
blv-assistive-vision-unified-viUnified_Dataset_with_EmotionsUnifiedBench
UnifiedBench
Dataset Summary
UnifiedBench is a curated benchmark of 800 harmful behavior prompts for evaluating the safety alignment of large language models (LLMs). It unifies and deduplicates entries from three widely used red-teaming datasets — AdvBench (Zou et al., 2023), HarmBench (Mazeika et al., 2024), and JailbreakBench (Chao et al., 2024) — under a single, consistent seven-category taxonomy anchored in the NIST AI 100-2 framework.
The dataset enables… See the full description on the dataset page: https://huggingface.co/datasets/labp2d/UnifiedBench.Unified_ReducedUnified_Reduced_All