Elamur Research⌕
☰
home / papers / 2026 / Reasoning Over Space: Enabling Geographic Reasoning for LLM-Based Generative Next POI Recommendation

Reasoning Over Space: Enabling Geographic Reasoning for LLM-Based Generative Next POI Recommendation

Paper · 2026 · Dongyi Lv, Qiuyu Ding, Heng-Da Xu et al.

Summary

The paper argues that LLM-based generative recommenders for next POI prediction make poor use of geography. Coordinates are usually added as prompt text or fused numeric features, so the model never reasons about distance or travel feasibility. The authors (Xi'an Jiaotong University and Amap, Alibaba Group) propose ROS (Reasoning Over Space), which has three parts:

  • Hierarchical Spatial Semantic ID (SID): an identifier built from an S2-cell location prefix, a semantic code for the POI category, and a suffix that keeps IDs unique.
  • Grounding pretraining: the model learns to map SIDs to and from text, and is trained on next-POI sequences.
  • Mobility Chain-of-Thought: a three-step procedure that profiles the user, builds a recency-ordered candidate set, and removes geographically implausible options. Spatially guided reinforcement learning then refines the model.

On three LBSN datasets (NYC, TKY, CA), ROS improves hit rate by more than 10% relative to the strongest LLM-based baselines, despite a smaller backbone. It also transfers better across cities and makes predictions with smaller distance errors. This matters for mobility and local-services applications, where geographically implausible suggestions are costly.

Key points

  • SID design: S2 Cell Id hex digits form coarse-to-fine location tokens, so nearby POIs share prefixes. Category descriptions are embedded with Qwen-0.6B and quantized with RQ-VAE.
  • RL stage: uses GRPO with three rewards (format, SID-level, and spatial). Predicting the correct coarse region earns partial credit even when the exact POI is wrong.
  • Top-K results on NYC: ROS reaches HR@5 = 0.6372 and HR@20 = 0.7284, against 0.6107 and 0.6977 for the best baseline. Baselines include SASRec, TIGER, GNPR-SID and OneLoc.
  • CoT-free variant (ROS†): it still beats all baselines, with HR@1 of 0.4181 (NYC), 0.3527 (TKY) and 0.2874 (CA). The pretrain-only model scores 0.3918, 0.3372 and 0.2657.
  • Limitation and code: the model assumes long trips within short intervals are unlikely, so it may undervalue fast travel by metro or highway. Code is at github.com/amap-cvlab/Reasoning-Over-Space.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🏢Alibaba GroupHangzhou · China168 works · 15,386 citations🎓Xi'an Jiaotong UniversityXi'an · China22 works · 36,046 citations

 Fields, topics and concepts

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksSocial Sciences · 1,772 worksInformation Systems · 1,437 worksSignal Processing · 452 worksTransportation · 6 works

🧭 Topics 1Research subjects and languages the work is about.Reasoning · 1,426 works

Cite

Cite as Dongyi Lv, Qiuyu Ding, Heng-Da Xu, Zhaoxu Sun, Zhi Wang, Feng Xiong and Mu Xu. 2026. Reasoning Over Space: Enabling Geographic Reasoning for LLM-Based Generative Next POI Recommendation. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.acl-long.332

@inproceedings{lv2026reasoning,
  title = {{Reasoning Over Space: Enabling Geographic Reasoning for LLM-Based Generative Next POI Recommendation}},
  author = {Dongyi Lv, Qiuyu Ding, Heng-Da Xu, Zhaoxu Sun, Zhi Wang, Feng Xiong and Mu Xu},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-long.332},
  url = {https://doi.org/10.18653/v1/2026.acl-long.332}
}

References

48 references, 3 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Junyi Chen, Lu Chi, Bingyue Peng, and Zehuan Yuan. 2024. Hllm: Enhancing sequential recommendations via hierarchical large language models for item and user modeling. arXiv preprint arXiv:2409.12740. — [arXiv:2409.12740](https://arxiv.org/abs/2409.12740) find
  2. Eunjoon Cho, Seth A. Myers, and Jure Leskovec. 2011. Friendship and mobility: user movement in location-based social networks. In Proceedings of the 17th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD ’11, page 1082–1090, New York, NY, USA. Association for Computing… find
  3. Jiaxin Deng, Shiyao Wang, Kuo Cai, Lejian Ren, Qigen Hu, Weifeng Ding, Qiang Luo, and Guorui Zhou. 2025. Onerec: Unifying retrieve and rank with generative recommender and iterative preference alignment. arXiv preprint arXiv:2502.18965. — [arXiv:2502.18965](https://arxiv.org/abs/2502.18965) find
  4. Shanshan Feng, Xutao Li, Yifeng Zeng, Gao Cong, Yeow Meng Chee, and Quan Yuan. 2015. Personalized ranking metric embedding for next new poi recommendation. In IJCAI, volume 15, pages 2069– 2075. find
  5. Shanshan Feng, Haoming Lyu, Fan Li, Zhu Sun, and Caishun Chen. 2024a. Where to move next: Zero-shot generalization of llms for next poi recommendation. In 2024 ieee conference on artificial intelligence (cai), pages 1530–1535. IEEE. find
  6. Shanshan Feng, Feiyu Meng, Lisi Chen, Shuo Shang, and Yew Soon Ong. 2024b. Rotan: A rotation-based temporal attention network for time-specific next poi recommendation. In Proceedings of the 30th ACM SIGKDD conference on knowledge discovery and data mining, pages 759–770. find
  7. Qiang Gao, Wei Wang, Kunpeng Zhang, Xin Yang, Congcong Miao, and Tianrui Li. 2022. Self-supervised representation learning for trip recommendation. Knowledge-Based Systems, 247:108791. find
  8. Shen Gao, Jiabao Fang, Quan Tu, Zhitao Yao, Zhumin Chen, Pengjie Ren, and Zhaochun Ren. 2024. Generative news recommendation. In Proceedings of the ACM Web Conference 2024, pages 3444–3453. find
  9. Marta C Gonzalez, Cesar A Hidalgo, and Albert-Laszlo Barabasi. 2008. Understanding individual human mobility patterns. nature, 453(7196):779–782. find
  10. Google Inc. 2023. S2 geometry library. https://s2geometry.io/. Accessed: 2025-03-01. ↗
  11. Sajal Halder, Kwan Hui Lim, Jeffrey Chan, and Xiuzhen Zhang. 2022. Efficient itinerary recommendation via personalized poi selection and pruning. Knowledge and Information Systems, 64(4):963–993. find
  12. Ruidong Han, Bin Yin, Shangyu Chen, He Jiang, Fei Jiang, Xiang Li, Chi Ma, Mincong Huang, Xiaoguang Li, Chunzhen Jing, and 1 others. 2025. Mtgr: Industrial-scale generative recommendation framework in meituan. In Proceedings of the 34th ACM International Conference on Information and Knowledge Manag… find
  13. Balázs Hidasi, Alexandros Karatzoglou, Linas Baltrunas, and Domonkos Tikk. 2015. Session-based recommendations with recurrent neural networks. arXiv preprint arXiv:1511.06939. — [arXiv:1511.06939](https://arxiv.org/abs/1511.06939) find
  14. Sepp Hochreiter and Jürgen Schmidhuber. 1997. Long short-term memory. Neural computation, 9(8):1735– 1780. find
  15. Guangneng Hu, Yu Zhang, and Qiang Yang. 2018. Mtnet: a neural approach for cross-domain recommendation with unstructured text. KDD deep learning day, pages 1–10. find
  16. Hao Jiang, Guoquan Wang, Donglin Zhou, Sheng Yu, Yang Zeng, Wencong Zeng, Kun Gai, and Guorui Zhou. 2025. Llm-aligned geographic item tokenization for local-life recommendation. arXiv preprint arXiv:2511.14221. — [arXiv:2511.14221](https://arxiv.org/abs/2511.14221) find
  17. Wang-Cheng Kang and Julian McAuley. 2018. Self-attentive sequential recommendation. In 2018 IEEE international conference on data mining (ICDM), pages 197–206. IEEE. find
  18. Doyup Lee, Chiheon Kim, Saehoon Kim, Minsu Cho, and Wook-Shin Han. 2022. Autoregressive image generation using residual quantization. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 11523–11532. find
  19. Sunkyung Lee, Minjin Choi, Eunseong Choi, Hye young Kim, and Jongwuk Lee. 2025. GRAM: Generative recommendation via semantic-aware multi-granular late fusion. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (ACL), pages 33294–33312. find
  20. Dapeng Li. 2018. 1.08 - geocoding and reverse geocoding. In Bo Huang, editor, Comprehensive Geographic Information Systems, pages 95–109. Elsevier, Oxford. find
  21. Peibo Li, Maarten de Rijke, Hao Xue, Shuang Ao, Yang Song, and Flora D Salim. 2024. Large language models for next point-of-interest recommendation. In Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 1463–1472. find
  22. Haitao Lin, Zhen Yang, Jiawei Xue, Ziji Zhang, Luzhu Wang, Yikun Gu, Yao Xu, and Xin Li. 2025. Spacetime-gr: A spacetime-aware generative model for large scale online poi recommendation. Preprint, arXiv:2508.16126. — [arXiv:2508.16126](https://arxiv.org/abs/2508.16126) find
  23. Zhao Liu, Wei Liu, Huajie Zhu, Jianxing Yu, Jian Yin, Wang-Chien Lee, and Shun Wang. 2025. Geography-aware large language models for next poi recommendation. arXiv preprint arXiv:2505.13526. — [arXiv:2505.13526](https://arxiv.org/abs/2505.13526) find
  24. Yan Luo, Haoyi Duan, Ye Liu, and Fu-Lai Chung. 2023. Timestamps as prompts for geography-aware location recommendation. In Proceedings of the 32nd ACM International Conference on Information and Knowledge Management, pages 1697–1706. find
  25. Yingtao Luo, Qiang Liu, and Zhaocheng Liu. 2021. Stan: Spatio-temporal attention network for next location recommendation. In Proceedings of the web conference 2021, pages 2177–2185. find
  26. Haohao Qu, Wenqi Fan, Zihuai Zhao, and Qing Li. 2025. Tokenrec: Learning to tokenize id for llm-based generative recommendations. IEEE Transactions on Knowledge and Data Engineering. find
  27. Shashank Rajput, Nikhil Mehta, Anima Singh, Raghu-nandan Hulikal Keshavan, Trung Vu, Lukasz Heldt, Lichan Hong, Yi Tay, Vinh Tran, Jonah Samost, and 1 others. 2023. Recommender systems with generative retrieval. Advances in Neural Information Processing Systems, 36:10299–10315. find
  28. Lütfi Kerem Senel, Besnik Fetahu, Davis Yoshida, Zhiyu Chen, Giuseppe Castellucci, Nikhita Vedula, Jason Ingyu Choi, and Shervin Malmasi. 2024. Generative explore-exploit: Training-free optimization of generative recommender systems using llm optimizers. In Proceedings of the 62nd Annual Meeting of… find
  29. Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, YK Li, Yang Wu, and 1 others. 2024. Deepseekmath: Pushing the limits of mathematical reasoning in open language models. arXiv preprint arXiv:2402.03300.
  30. Roger W Sinnott. 1984. Virtues of the haversine. Sky and telescope, 68(2):158. find
  31. Fei Sun, Jun Liu, Jian Wu, Changhua Pei, Xiao Lin, Wenwu Ou, and Peng Jiang. 2019. Bert4rec: Sequential recommendation with bidirectional encoder representations from transformer. In Proceedings of the 28th ACM international conference on information and knowledge management, pages 1441–1450. find
  32. Jiaxi Tang and Ke Wang. 2018. Personalized top-n sequential recommendation via convolutional sequence embedding. In Proceedings of the eleventh ACM international conference on web search and data mining, pages 565–573. find
  33. Alicia Tsai, Adam Kraft, Long Jin, Chenwei Cai, Anahita Hosseini, Taibai Xu, Zemin Zhang, Lichan Hong, Ed H Chi, and Xinyang Yi. 2024. Leveraging llm reasoning enhances personalized recommender systems. In Findings of the Association for Computational Linguistics: ACL 2024, pages 13176–13188. find
  34. Dongsheng Wang, Yuxi Huang, Shen Gao, Yifan Wang, Chengrui Huang, and Shuo Shang. 2025. Generative next poi recommendation with semantic id. In Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V. 2, pages 2904–2914. find
  35. Lei Wang, Wanyu Xu, Yihuai Lan, Zhiqiang Hu, Yunshi Lan, Roy Ka-Wei Lee, and Ee-Peng Lim. 2023. Plan-and-solve prompting: Improving zero-shot chain-of-thought reasoning by large language models. arXiv preprint arXiv:2305.04091. find
  36. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, and 1 others. 2022. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 35:24824– 24837.
  37. Zhipeng Wei, Kuo Cai, Junda She, Jie Chen, Minghao Chen, Yang Zeng, Qiang Luo, Wencong Zeng, Ruiming Tang, Kun Gai, and 1 others. 2025. Oneloc: Geo-aware generative recommender systems for local life service. arXiv preprint arXiv:2508.14646. — [arXiv:2508.14646](https://arxiv.org/abs/2508.14646) find
  38. Xiaodong Yan, Tengwei Song, Yifeng Jiao, Jianshan He, Jiaotuan Wang, Ruopeng Li, and Wei Chu. 2023. Spatio-temporal hypergraph learning for next poi recommendation. In Proceedings of the 46th international ACM SIGIR conference on research and development in information retrieval, pages 403–412. find
  39. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, and 1 others. 2025. Qwen3 technical report. arXiv:2505.09388. arXiv preprint Dingqi Yang, Daqing Zhang, Vincent W. Zheng, and Zhiyong Yu. 2015. Modeling user activity preference by…
  40. Song Yang, Jiamou Liu, and Kaiqi Zhao. 2022. Getnext: Trajectory flow map enhanced transformer for next poi recommendation. In Proceedings of the 45th International ACM SIGIR Conference on research and development in information retrieval, pages 1144– 1153. find
  41. Lih Wei Yeow, Raymond Low, Yu Xiang Tan, and Lynette Cheah. 2021. Point-of-interest (poi) data validation methods: An urban case study. ISPRS International Journal of Geo-Information, 10(11):735. find
  42. Weiqi Yue, Yuyu Yin, Xin Zhang, Binbin Shi, Tingting Liang, and Jian Wan. 2025. Cot4rec: Revealing user preferences through chain of thought for recommender systems. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 39, pages 13142–13151. find
  43. Jiaqi Zhai, Lucy Liao, Xing Liu, Yueming Wang, Rui Li, Xuan Cao, Leon Gao, Zhaojie Gong, Fangda Gu, Jiayuan He, Yinghai Lu, and Yu Shi. 2024. Actions speak louder than words: trillion-parameter sequential transducers for generative recommendations. In Proceedings of the 41st International Conference… find
  44. Penglong Zhai, Jie Li, Fanyi Di, Yue Liu, Yifang Yuan, Jie Huang, Peng Wu, Sicong Wang, Mingyang Yin, Tingting Hu, and 1 others. 2025. Cognitive-aligned spatio-temporal large language models for next point-of-interest prediction. arXiv preprint arXiv:2510.14702. — [arXiv:2510.14702](https://arxiv.org/abs/2510.14702) find
  45. Qianru Zhang, Peng Yang, Junliang Yu, Haixin Wang, Xingwei He, Siu-Ming Yiu, and Hongzhi Yin. 2025. A survey on point-of-interest recommendation: Models, architectures, and security. IEEE Transactions on Knowledge and Data Engineering. find
  46. Yu Zheng. 2015. Trajectory data mining: an overview. ACM Transactions on Intelligent Systems and Technology (TIST), 6(3):1–41. find
  47. Lin Zhong, Lingzhi Wang, Xu Yang, and Qing Liao. 2025. Comapoi: A collaborative multi-agent framework for next poi prediction bridging the gap between trajectory and language. In Proceedings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 17… find
  48. Kun Zhou, Hui Wang, Wayne Xin Zhao, Yutao Zhu, Sirui Wang, Fuzheng Zhang, Zhongyuan Wang, and Ji-Rong Wen. 2020. S3-rec: Self-supervised learning for sequential recommendation with mutual information maximization. In Proceedings of the 29th ACM international conference on information & knowledge man… find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/lv-2026-reasoning-over-space-enabling-geographic/