This dataset contains real and AI-generated image samples organized for binary MLLM-generated image detection.
Paper | Code
We construct an MLLM-generated image detection benchmark from GPT Image2 and Nano Banana2. This benchmark covers texture-dominated, structure-dominated, and hybrid-dominated. It is designed to evaluate detector performance under the new challenges introduced by large-scale image generation models.… See the full description on the dataset page:
https://huggingface.co/datasets/zr-zhang/MLLM-Generated-Image-Detection-Dataset.