datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
HaloQuest
HaloQuest
A visual question answering benchmark for evaluating hallucination in vision-language models. Contains 7,609 examples with both real (Open Images) and synthetic (Midjourney) images.
Fields
Field
Description
id
Row identifier
image
Input image
image_name
Original image filename
url
Source URL
image_type
real or generated
hallucination_type
false premises, visual challenge, or insufficient context
question
Question about the image… See the full description on the dataset page: https://huggingface.co/datasets/MM-Hallu/HaloQuest.halo-infinite-angel-videogame
Dataset Card for halo-infinite-angel-videogame
** The original COCO dataset is stored at dataset.tar.gz**
Dataset Summary
halo-infinite-angel-videogame
Supported Tasks and Leaderboards
object-detection: The dataset can be used to train a model for Object Detection.
Languages
English
Dataset Structure
Data Instances
A data point comprises an image and its object annotations.
{
'image_id': 15,
'image':… See the full description on the dataset page: https://huggingface.co/datasets/Francesco/halo-infinite-angel-videogame.halo-flare-dataset
HALO Synthetic Lens-Flare Dataset ✨
HALO is a 3D-rendered synthetic lens-flare dataset companion to our CVPR 2026 paper, UniSER: A Foundation Model for Unified Soft Effects Removal. Each sample is a Blender-rendered 4K scene with a paired flare-free / flared / flare-only triplet — designed for training and benchmarking single-image lens-flare removal.
📦 Size: ~150 GB / 4945 samples / ~80 WebDataset shards
📄 Paper: arXiv 2511.14183
🐙 Code:… See the full description on the dataset page: https://huggingface.co/datasets/jdzhang0929/halo-flare-dataset.HaloQuestThis Dataset introduces challenging pictures for assessing MLLM hallucinations. It includes questions and ground-truth answers.
Original Paper: (https://arxiv.org/abs/2407.15680)[https://arxiv.org/abs/2407.15680]
Code Repo: (https://github.com/google/haloquest/)[https://github.com/google/haloquest/]
halosight-dataghibli-dataset
Ghibli Real vs AI-Generated Dataset
One sample per line
Includes: id, image, label, description
Use this for standard classification or image-text training
Real images sourced from Nechintosh/ghibli (810 images)
AI-generated images created using:
nitrosocke/Ghibli-Diffusion (2637 images)
KappaNeuro/studio-ghibli-style (810 images)
Note: While the KappaNeuro repository does not explicitly state a license, it is a fine-tuned model based on Stable Diffusion XL, which is… See the full description on the dataset page: https://huggingface.co/datasets/Halo8024/ghibli-dataset.haloquest_testHaloQuesttwitter-Halou1114172-2025.08.29-1961272689475109347-ccoueWn_0nj_cy0s-part1