发表机构
ZODA; HKUST; UCAS; ZJU; UTokyo（ZODA; 香港科技大学; 中国科学院大学; 浙江大学; 东京大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
为解决视觉问题求解评估缺失，提出SolveEdit基准，含2,728个场景变换案例，并引入两阶段规划器SolveEdiT-PLAN，平均提升SolvEScore 9.1分。
AI 中文摘要
机器智能通常通过抽象推理问题进行评估，然而许多现实世界的问题是视觉性的，例如排列物体、修复布局或追踪路线。解决这些问题需要理解场景，推断为实现目标必须改变什么，并在不干扰无关内容的情况下实现该改变。然而，现有基准主要评估感知、生成或明确指定的变换，导致目标驱动的视觉问题求解未被充分探索。为弥补这一差距，我们引入了SolveEpIT，一个通过场景变换进行视觉问题求解的基准。给定一张图像和一个目标，模型必须从请求、场景或视觉表达的规则中推断出有效的变换，然后执行该变换，同时保留无关内容。SoLvEEDrr包含2,728个案例。原子变换契约规定了必需条件和保护条件，使SoLvEScoRE能够在没有单一参考输出的情况下衡量完成度和意外变化。评估的最强模型仅达到57.0%的SolvEScore。我们进一步引入了SolveEdiT-PLAN，一个两阶段视觉规划器，在生成前实例化变换。在匹配的单次生成评估下，它在三个测试生成器上平均提高了SoLvEScoRE 9.1个百分点，包括GPT-Image-2从57.0%提升至71.6%，而无需修改编辑器。
英文摘要
Machine intelligence is often evaluated through abstract reasoning problems, yet many real-world problems are visual, such as arranging objects, repairing layouts, or tracing routes. Solving these problems requires understanding a scene, inferring what must change to achieve a goal, and realizing that change without disturbing unrelated content. However, existing benchmarks mainly evaluate perception, generation, or explicitly specified transformations, leaving goal-driven visual problem solving underexplored. To bridge this gap, we introduce SolveEpIT, a benchmark for visual problem solving through scene transformation. Given an image and a goal, a model must infer a valid transformation from the request, the scene, or a visually expressed rule, then execute it while preserving unrelated content. SoLvEEDrr contains 2,728 cases. Atomic transition contracts specify required and protected conditions, enabling SoLvEScoRE to measure completion and unintended changes without a single reference output. The strongest evaluated model achieves only57.0% SolvEScore. We further introduce SolveEdiT-PLAN, a two-stage visual planner that instantiates the transition before generation. Under matched single-generation evaluation, it improves SoLvEScoRE by 9.1 points on average across three tested generators, including a gain from 57.0% to 71.6% for GPT-Image-2, without modifying the editor.
