Artificial Analysis 评测 Qwen-Image-2.1 图像编辑能力
Where Qwen-Image-2.1 is strongest in Image Editing: closest to the category frontier among all models in Scene & Style Edit, Text or Symbol Edits and Enhancemen
Artificial Analysis 的 Image Arena 分类体系按 7 种编辑动作和 10 类使用场景分别设立榜单，用于衡量图像编辑模型的表现。Qwen-Image-2.1 在 Scene & Style Edit 上最接近该类别前沿，其次是 Text or Symbol Edits 和 Enhancement & Restoration。相比 Qwen Image 2.0，它在全部 7 种编辑动作上都缩小了与前沿的差距，其中 Text or Symbol Edits、Object-Level Edit 和 Identity Preserving Edit 提升最大。
正文
当前显示原文，中文全文译文尚未就绪。
Where Qwen-Image-2.1 is strongest in Image Editing: closest to the category frontier among all models in Scene & Style Edit, Text or Symbol Edits and Enhancement & Restoration
Our Image Arena taxonomy measures 7 editing actions and 10 use cases for Image Editing, each with its own leaderboard. Editing actions measure what an edit asks the model to do.
Qwen-Image-2.1 is closest to the category frontier in Scene & Style Edit, followed by Text or Symbol Edits and Enhancement & Restoration. Against Qwen Image 2.0 it closes the gap to the frontier on all seven editing actions, with the largest gains in Text or Symbol Edits, Object-Level Edit and Identity Preserving Edit.
➤ Scene & Style Edit covers relighting, restyling, and changes to the background, setting, or overall treatment.
➤ Text or Symbol Edits covers editing text, typography, logos, and symbols within the image.
➤ Enhancement & Restoration covers denoising, retouching, artifact removal, and repair of damaged or degraded images.