GUI-Primitives: Diagnosing Spatial Reasoning Failures in Vision-Language GUI Grounding
- 类型：arxiv
- 标识：2608.21832
- 链接：https://arxiv.org/abs/2608.21832
- 主分类：multimodal
- 形态：benchmark
- 被引：0
- 被引来源：Semantic Scholar + OpenAlex
- S2被引：0
- OpenAlex被引：0
- 影响力被引：0
- TLDR：Computer-use agents ground natural-language instructions in screenshots to locate interface elements, yet existing benchmarks do not isolate whether models bind relational language to the correct element. We introduce GUI-Primitives, a 994-item benchmark of contrastive instruction pairs over seven spatial relations in graphical user interfaces (left/right, above/below, containment, alignment, proximity, list ordinal, occlusion). Each pair holds the screenshot and anchor fixed while changing the relation expression, so the correct target moves between two designated candidates. Five annotators
- OpenAlex ID：W7204194524
- OpenAlex DOI：10.48550/arxiv.2608.21832
- DOI：10.48550/arxiv.2608.21832
- DOI来源：OpenAlex
- 开放获取：green
- 开放获取链接：https://arxiv.org/pdf/2608.21832
- OpenAlex更新：2026-09-01
- 副分类：evaluation
- 待LLM分类：否
- 标题中文：GUI-Primitives：诊断视觉语言 GUI grounding 中的空间推理失败
- TLDR中文：计算机使用 Agent 将自然语言指令 grounding 到截图中以定位界面元素，但现有基准无法隔离模型是否将关系语言正确绑定到对应元素。我们提出 GUI-Primitives，一个包含 994 个条目的基准，由跨七种空间关系（左右、上下、包含、对齐、邻近、列表序数、遮挡）的对比指令对组成。每对保持截图和锚点不变，仅改变关系表达，使正确目标在两个指定候选之间切换。五位标注者……
- 来源文件：
- /inbox/tom/_candidates/2026-08-28-agent-rag-longcontext-candidates.json
- [S2 enrich]
- [OpenAlex backfill]