arXiv cs.CL· Mifu Matsumoto, Yuki Nakanishi, Tomoya Hashiguchi, Hiroaki Ohshima·· 11 小时前AI 评分24
用 LLM 生成地标路线指引：基于 SVG 结构化地图表示的研究
Structured Maps for Landmark-Based Route Instructions with LLMs
AI 导读
研究提出用 SVG 格式的结构化文本编码地图空间特征，让 LLM 直接获取几何与语义数据以完成地标选择任务。基于人类路线描述与 OpenStreetMap 数据构建的新数据集上，五个 LLM 的测试显示该结构化文本输入效果与 PNG 图像输入相当或更好。任务特定微调还能让较小的本地 LLM 利用地图布局达到高地标选择准确率。
正文 · 原文
Abstract:To generate landmark-based route instructions, models must understand both map geometry and human landmark preferences. Although large language models (LLMs) show great abilities in language tasks, the best way to structure map data for these models is not yet established. This paper proposes a structured text representation for landmark selection tasks in route navigation, utilizing SVG format to encode spatial features. The structured text maintains a geometric connection with rendered map images, allowing language models to directly access spatial and semantic data. Using a new dataset created from human-written route descriptions and OpenStreetMap data, we evaluate models on two tasks: landmark selection at turns and around destinations. Experiments with five LLMs show that the proposed structured text input performs as well as or better than image-based PNG inputs. In addition, task-specific fine-tuning enables smaller local LLMs to effectively leverage map layouts to achieve high selection accuracy. These results demonstrate that structured text is an efficient and lightweight interface for spatial reasoning in human-oriented navigation.
| Comments: | Accepted for publication at BESC 2026 |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.36582 [cs.CL] |
| (or arXiv:2610.36582v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.36582 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Hiroaki Ohshima [view email]
[v1]
Tue, 27 Oct 2026 03:36:49 UTC (369 KB)
来源：arXiv cs.CL · arxiv.org