UCSC-VLAA/Complex-Edit
收藏Hugging Face2025-07-31 更新2025-05-31 收录
下载链接：
https://hf-mirror.com/datasets/UCSC-VLAA/Complex-Edit
下载链接
链接失效反馈官方服务：
资源简介：
Complex-Edit是一个全面的基准测试，旨在系统地评估基于指令的图像编辑模型在不同复杂度指令下的表现。该数据集利用GPT-4o自动收集了多样化的编辑指令，并通过一系列指标和基于VLM的自动评估管道来支持大规模评估。
Complex-Edit is a comprehensive benchmark designed to systematically evaluate instruction-based image editing models across instructions of varying complexity. The dataset leverages GPT-4o to automatically collect a diverse set of editing instructions and is supported by a suite of metrics and a VLM-based auto-evaluation pipeline for large-scale assessments.
提供机构：
UCSC-VLAA
