← 研踪 PaperTrailCVF · CVPR 2026:cvf:69d8f4e5b9c4c357d42a4d11613748d0 · 记录 #1
CompBench: Benchmarking Complex Instruction-guided Image Editing
Jia, BohanHuang, WenxuanTang, YuntianQiao, JunboLiao, JinchengCao, ShaoshengZhao, FeiFeng, ZhaopengGu, ZhouhongYin, ZhenfeiBai, LeiOuyang, WanliChen, LinZhao, FeiWang, ZihanXie, YuanLin, Shaohui 来源日期（仅到年份） 2026
来源记录日期 2026
雷达观测 2026-11-01 16:37
摘要
While real-world applications increasingly demand intricate scene manipulation, existing instruction-guided image editing benchmarks often oversimplify task complexity and lack comprehensive, fine-grained instructions. To bridge this gap, we introduce CompBench, a large-scale benchmark specifically designed for complex instruction-guided image editing. CompBench features challenging editing scenarios that incorporate fine-grained instruction following, spatial and contextual reasoning, thereby enabling comprehensive evaluation of image editing models' precise manipulation capabilities. To construct CompBench, we propose an MLLM-human collaborative framework with tailored task pipelines. Furthermore, we propose an instruction decoupling strategy that disentangles editing intents into four key dimensions: location, appearance, dynamics, and objects, ensuring closer alignment between instructions and complex editing requirements. Extensive evaluations reveal that CompBench exposes fundamental limitations of current image editing models and provides critical insights for the development of next-generation instruction-guided image editing systems. Our project page is available at https://comp-bench.github.io/.
论文发现出处
这些页面提供论文线索，不代表同行评审或质量认证。
观测到的版本
平台记录号代表本站观测到的元数据变化，不是作者修订号。来源日期仅到月份或年份时，按该月或该年首日存档。
记录 #1 · 2026 · 最新观测版本
CompBench: Benchmarking Complex Instruction-guided Image Editing
While real-world applications increasingly demand intricate scene manipulation, existing instruction-guided image editing benchmarks often oversimplify task complexity and lack comprehensive, fine-grained instructions. To bridge this gap, we introduce CompBench, a large-scale benchmark specifically designed for complex instruction-guided image editing. CompBench features challenging editing scenarios that incorporate fine-grained instruction following, spatial and contextual reasoning, thereby enabling comprehensive evaluation of image editing models' precise manipulation capabilities. To construct CompBench, we propose an MLLM-human collaborative framework with tailored task pipelines. Furthermore, we propose an instruction decoupling strategy that disentangles editing intents into four key dimensions: location, appearance, dynamics, and objects, ensuring closer alignment between instructions and complex editing requirements. Extensive evaluations reveal that CompBench exposes fundamental limitations of current image editing models and provides critical insights for the development of next-generation instruction-guided image editing systems. Our project page is available at https://comp-bench.github.io/.
查看此版本解读 →查看此版本原文 ↗