datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
CLEAR-Bias
Dataset Card for CLEAR-Bias
CLEAR-Bias (Corpus for Linguistic Evaluation of Adversarial Robustness against Bias) is a benchmark dataset designed to assess the robustness of large language models (LLMs) against bias elicitation, especially under adversarial conditions.
It consists of carefully curated prompts that test the ability of LLMs to resist generating biased content when exposed to both standard and adversarial inputs.
The dataset targets a broad spectrum of social biases… See the full description on the dataset page: https://huggingface.co/datasets/RCantini/CLEAR-Bias.rpg_fantasy_battle_counterfactual_v2rpg_fantasy_battle_counterfactual_v3rc-annotated-sftr-canada-shrunktest_workspacerpg_fantasy_battleDataset of images, text captions (and their CLIP embeddings), and variables for a fantasy RPG battle scene.
Each row in the dataset contains an image, a list of captions describing the image (and their CLIP embeddings), and a list of variables describing the scene.
The images are in PIL.Image.Image format, captions are a list of strings, CLIP embeddings are a tensor, and the variables are in a Python dictionary.
The dataset is intended for use in training models for generating counterfactuals… See the full description on the dataset page: https://huggingface.co/datasets/rcannizzaro/rpg_fantasy_battle.rpg_fantasy_battle_partial_descriptionsrpg_fantasy_battle_counterfactual