HuggingFace Daily Papers·· 1 天前
EditAgent：重新思考智能体图像编辑
EditAgent: Rethinking Agentic Image Editing
AI 导读
EditAgent 通过扩展智能体的动作空间与观察空间改进图像编辑：每一步同时路由到通用编辑器与专用方法，并为每个候选生成验证记录。在专家标注基准上，通用编辑器超 40% 的案例无可用结果，而现成 VLM 评判器借助验证记录可超过微调奖励模型。候选并行生成、一次 listwise 评判，专家在 78.5% 的有效比较中更偏好该方式，且评判调用不超过一半。
正文 · 原文
Abstract:Agentic image editors decompose complex requests into steps, but when a step fails, they mostly retry with the same narrow set of tools and a judge that scores each output by eye directly. We argue that such agents are bounded by two spaces: the action space of what the agent can do, and the observation space of what its judge can see. On our expert-annotated benchmark, a general-purpose editor yields no usable result in over 40% of the cases, and no judge is reliable across edit types. We present EditAgent, which expands both spaces. It routes each step to general editors and specialized methods, and attaches to each candidate a verification record. With these records, off-the-shelf VLM judges outperform finetuned reward models. Candidates are generated in parallel and judged in one listwise call: experts prefer this to sequential refinement in 78.5% of decided comparisons, at no more than half the judge calls. EditAgent improves two open-weight editors on three benchmarks and serves as a data engine for training stronger editors.
| Subjects: | Computer Vision and Pattern Recognition (cs.CV) |
| Cite as: | arXiv:2610.39664 [cs.CV] |
| (or arXiv:2610.39664v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2610.39664 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Haiyang Xu [view email]
[v1]
Thu, 29 Oct 2026 09:15:52 UTC (36,003 KB)
来源：HuggingFace Daily Papers · arxiv.org