wj-inf/MagicAssessor-7B

MagicAssessor-7B

MagicAssessor-7B is a Vision-Language Model (VLM) developed for fine-grained artifact assessment in text-to-image generation. It is a core component of the comprehensive MagicMirror framework, which aims to systematically evaluate the perceptual quality and identify various anatomical and structural flaws in generated images.

The model was introduced in the paper MagicMirror: A Large-Scale Dataset and Benchmark for Fine-Grained Artifacts Assessment in Text-to-Image Generation.

• —Paper: arXiv:2509.10260 | Hugging Face Papers: 2509.10260
• —Project Page: https://wj-inf.github.io/MagicMirror-page/
• —Code / GitHub Repository (MagicMirror Benchmark): https://github.com/wj-inf/MagicMirror
• —Dataset (MagicData340K): https://huggingface.co/datasets/wj-inf/MagicData340k
• —Model (MagicAssessor-7B - this repository): https://huggingface.co/wj-inf/MagicAssessor-7B