WeGenEditBench 图像编辑基准发布：GPT-Image-2 与 Seedream-5.0-Pro 分项领跑
From Evaluation to Evolution: Benchmarking Image Editing and Evolving Image Models
一篇把图像编辑模型逐项挑毛病的论文，还配了能自动诊断失败并针对性改进的框架，FLUX.2 从 1 分多提到 3.5 分，做图像生成的可以细看。
论文提出图像编辑基准 WeGenEditBench，包含 2,740 个评估实例和 148 种任务类型，覆盖单图、文本密集和多参考图编辑三类场景。评估采用预生成检查清单，把每个请求拆解为原子需求，逐项扣分并附视觉证据，汇总为执行、保真、画质和任务正确性四类得分。15 个受测配置中，GPT-Image-2 以 4.076 分领先单图编辑，Seedream-5.0-Pro 以 4.291 分领先文本密集编辑。配套的 WeGenEvoStudio 多智能体框架面向递归自我改进（RSI），实验中 FLUX.2 [dev] 在 Top-5 单图目标上总分从 1.083 提升到 3.517，Qwen-Image-Edit-2511 提升 13.2%。
From Evaluation to Evolution: Benchmarking Image Editing and Evolving Image Models
Real-world image editing requires coordinating multiple visual transformations with content preservation and reference constraints, yet aggregate scores can readily obscure unmet local requirements. Effective evaluation must therefore identify local failures and translate them into priorities for subsequent model improvement. We introduce WeGenEditBench, comprising 2,740 evaluation instances and 148 task types across single-image, text-intensive, and multi-reference editing. Pre-generated checklists shared across models decompose each request into atomic requirements; item-level deductions and visual evidence are established first and then summarized into scores for edit execution, preservation, visual quality, and task-specific correctness. To support the experience-acquisition component of recursive self-improvement (RSI), we develop WeGenEvoStudio, a unified multi-agent harness for RSI-oriented learning that integrates model-conditioned failure diagnosis, retrieval-grounded data construction, supervised adaptation, and re-evaluation. Among 15 evaluated configurations, GPT-Image-2 leads single-image editing with an Overall score of 4.076, while Seedream-5.0-Pro leads text-intensive editing at 4.291. Evolution experiments across four generation and editing benchmarks demonstrate substantial targeted gains: on their report-selected Top-5 single-image targets, FLUX.2 [dev] improves Overall from 1.083 to 3.517, while Qwen-Image-Edit-2511 gains 13.2%. Together, our benchmark and harness provide a foundation for developing more reliable image models through interpretable evaluation and feedback-driven learning.