Language Modeling for Urban Mobility: A Data-Centric Review and Guidelines.
This repository contains a collection of resources and papers on applying language modeling paradigms to urban mobility scenarios. If you find this repo is useful, please cite our paper [bib].
We propose a comprehensive and data-centric survey of language modeling for urban mobility, structured along three key dimensions:
- (i) How to transform heterogeneous mobility data into language model–like formats through tokenization, encoding, and prompting;
- (ii) How to choose among different categories of language models, ranging from pretrained language models, large language models (LLMs), MLLMs, and diffusion models;
- (iii) What are the advantages of applying language modeling to urban mobility in diverse urban downstream tasks?
- [Science 2010] Network Diversity and Economic Development [paper]
- [Science 2015] Predicting poverty and wealth from mobile phone metadata [paper]
- [Nature 2021] The universal visitation law of human mobility [paper]
- [Nature 2022] Machine learning and phone data can improve targeting of humanitarian aid [paper]
- [Nature Machine Intelligence 2022] Quantifying the spatial homogeneity of urban road networks via graph neural networks [paper]
- [Nature Computational Science 2023] Future directions in human mobility science [paper]
- [Nature Communications 2024] Similarity and economy of scale in urban transportation networks and optimal transport-based infrastructures [paper]
- [Nature Communications 2024] Unravelling the spatial directionality of urban mobility [paper]
- [Nature Cities 2024] Infrequent activities predict economic outcomes in major American cities [paper]
- [HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS 2024] Neural embeddings of urban big data reveal spatial structures in cities [paper]
- [HUMANITIES AND SOCIAL SCIENCES COMMUNICATIONS 2024] Counterfactual mobility network embedding reveals prevalent accessibility gaps in U.S. cities [paper]
- [Nature Cities 2024] Network constraints on worker mobility [paper]
- [Nature Human Behaviour 2025] Using human mobility data to quantify experienced urban inequalities [paper]
- [Nature Human Behaviour 2025] Behaviour-based dependency networks between places shape urban economic resilience [paper]
- [Nature Communications 2025] Human mobility is well described by closed-form gravity-like models learned automatically from data [paper]
-
A survey on deep learning for human mobility (2021). [paper] [code]
- Massimiliano Luca, Gianni Barlacchi, Bruno Lepri, Luca Pappalardo
-
MobilityDL: a review of deep learning from trajectory data (2025). [paper] [code]
- Anita Graser, Anahid Jalali, Jasmin Lampert, Axel Weißenfeld, Krzysztof Janowicz
-
Deep learning for trajectory data management and mining: A survey and beyond (2024). [paper] [code]
- Wei Chen, Yuxuan Liang, Yuanshao Zhu, Yanchuan Chang, Kang Luo, Haomin Wen, Lei Li, Yanwei Yu, Qingsong Wen, Chao Chen, et al.
-
A survey of large language models for traffic forecasting: Methods and applications (2025). [paper]
- Qingqing Long, Shuai Liu, Ning Cao, Zhicheng Ren, Wei Ju, Chen Fang, Zhihong Zhu, Hengshu Zhu, Yuanchun Zhou
-
Large language models for mobility analysis in transportation systems: A survey on forecasting tasks (2024). [paper]
- Zijian Zhang, Yujie Sun, Zepu Wang, Yuqi Nie, Xiaobo Ma, Ruolin Li, Peng Sun, Xuegang Ban
-
Large Language Models for Urban Mobility (2025). [paper]
- Youssef Hussein, Mohamed Hemdan, Mohamed F Mokbel
-
Large models for time series and spatio-temporal data: A survey and outlook (2023). [paper] [code]
- Ming Jin, Qingsong Wen, Yuxuan Liang, Chaoli Zhang, Siqiao Xue, Xue Wang, James Zhang, Yi Wang, Haifeng Chen, Xiaoli Li, et al.
-
Foundation models for spatio-temporal data science: A tutorial and survey (2025). [paper]
- Yuxuan Liang, Haomin Wen, Yutong Xia, Ming Jin, Bin Yang, Flora Salim, Qingsong Wen, Shirui Pan, Gao Cong
-
Unraveling Spatio-Temporal Foundation Models via the Pipeline Lens: A Comprehensive Review (2025). [paper] [code]
- Yuchen Fang, Hao Miao, Yuxuan Liang, Liwei Deng, Yue Cui, Ximu Zeng, Yuyang Xia, Yan Zhao, Torben Bach Pedersen, Christian S Jensen, et al.
-
How can time series analysis benefit from multiple modalities? a survey and outlook (2025). [paper] [code]
- Haoxin Liu, Harshavardhan Kamarthi, Zhiyuan Zhao, Shangqing Xu, Shiyu Wang, Qingsong Wen, Tom Hartvigsen, Fei Wang, B Aditya Prakash
-
Multi-modal time series analysis: A tutorial and survey (2025). [paper] [code]
- Yushan Jiang, Kanghui Ning, Zijie Pan, Xuyang Shen, Jingchao Ni, Wenchao Yu, Anderson Schneider, Haifeng Chen, Yuriy Nevmyvaka, Dongjin Song
-
Urban computing in the era of large language models (2025). [paper] [code]
- Zhonghang Li, Lianghao Xia, Xubin Ren, Jiabin Tang, Tianyi Chen, Yong Xu, Chao Huang
-
Towards urban general intelligence: A review and outlook of urban foundation models (2024). [paper]
- Weijia Zhang, Jindong Han, Zhao Xu, Hang Ni, Hao Liu, Hui Xiong
-
Attention
-
Masked LM
- CTLE (
lin2021pre) Pre-training context and time aware location embeddings from spatial-temporal trajectories for user next location prediction. AAAI, 2021. [paper] - Wepos (
guo2022wepos) Wepos: Weak-supervised indoor positioning with unlabeled wifi for on-demand delivery. IMWUT, 2022. [paper] - Yang et al. (
yang2024applying) Applying masked language model for transport mode choice behavior prediction. TR-A, 2024. [paper] - GREEN (
zhou2025grid) Grid and road expressions are complementary for trajectory representation learning. KDD, 2025. [paper] [code] - TrajBERT (
trajbert2023) TrajBERT: BERT-Based Trajectory Recovery with Spatial-Temporal Refinement. TMC, 2023. [paper]
- CTLE (
-
Transformer Decoder
- MobilityGPT (
mobilitygpt2024) MobilityGPT: Enhanced Human Mobility Modeling with a GPT model. [paper] [code] - GeoFormer (
solatorio2023geoformer) GeoFormer: predicting human mobility using generative pre-trained transformer (GPT). 2023. [paper] [code] - LMTAD (
mbuya2024trajectory) Trajectory Anomaly Detection with Language Models. SIGSPATIAL, 2024. [paper] [code] - Kobayashi et al. (
kobayashi2023modeling) Modeling and generating human mobility trajectories using transformer with day encoding. 2023. [paper] - Traj-LLM (
lan2024traj) Traj-llm: A new exploration for empowering trajectory prediction with pre-trained large language models. IEEE Transactions on Intelligent Vehicle, 2024. [paper] - TrajLearn (
nadiri2025trajlearn) TrajLearn: Trajectory Prediction Learning using Deep Generative Models. ACM Transactions on Spatial Algorithms and Systems, 2025. [paper] [code]
- MobilityGPT (
- GNPR-SID (
wang2025generativekdd25) Generative Next POI Recommendation with Semantic ID. KDD, 2025. [paper] [code] - RHYTHM (
he2025rhythm) RHYTHM: Reasoning with Hierarchical Temporal Tokenization for Human Mobility. arXiv, 2025. [paper] [code] - MobGLM (
zhang2024mobglm) MobGLM: A Large Language Model for Synthetic Human Mobility Generation. SIGSPATIAL, 2024. [paper] - MobilityGPT (
mobilitygpt2024) MobilityGPT: Enhanced Human Mobility Modeling with a GPT model. [paper] [code] - Geo-Llama (
li2024geo) Geo-llama: Leveraging llms for human mobility trajectory generation with spatiotemporal constraints. MDM, 2025. [paper] [code]
- Deepmove (
feng2018deepmove) DeepMove: Predicting human mobility with attentional recurrent networks. WWW, 2018. [paper] - STRNN (
liu2016predicting) Predicting the next location: A recurrent model with spatial and temporal contexts. AAAI, 2016. [paper] - LSTPM (
sun2020go) Where to go next: Modeling long-and short-term user preferences for point-of-interest recommendation. AAAI, 2020. [paper] - STAN (
Luo2021stan) STAN: Spatio-Temporal Attention Network for Next Location Recommendation. WWW, 2021. [paper] [code]
-
Masked LM
- LP-BERT (
suzuki2024cross) Cross-city-aware Spatiotemporal BERT. SIGSPATIAL, 2024. [paper] - TraceBERT (
crivellari2022tracebert) Tracebert—a feasibility study on reconstructing spatial--temporal gaps from incomplete motion trajectories via bert training process on discrete location sequences. Sensors, 2022. [paper] - CTLE (
lin2021pre) Pre-training context and time aware location embeddings from spatial-temporal trajectories for user next location prediction. AAAI, 2021. [paper] - GREEN (
zhou2025grid) Grid and road expressions are complementary for trajectory representation learning. KDD, 2025. [paper] [code]
- LP-BERT (
-
Causal Attention
- LLMEmb (
liu2025llmemb) Llmemb: Large language model can be a good embedding generator for sequential recommendation. AAAI, 2025. [paper] [code] - Mobility-LLM (
mobilityllm2024) Mobility-llm: Learning visiting intentions and travel preference from human mobility data with large language models. NeurIPS, 2024. [paper] [code] - NextLocLLM (
nextlocllm2025) NextlocLLM: Next Location Prediction Using LLMs. arXiv, 2025. [paper] [code] - GSTM-HMU (
luo2025gstm) GSTM-HMU: Generative Spatio-Temporal Modeling for Human Mobility Understanding. arXiv, 2025. [paper]
- Cardiff (
guo2025leveraging) Leveraging the Spatial Hierarchy: Coarse-to-fine Trajectory Generation via Cascaded Hybrid Diffusion. arXiv preprint arXiv:2507.13366, 2025. [paper] [code] - Diff-POI (
qin2023diffusion) A Diffusion Model for POI Recommendation. ACM Transactions on Information Systems, 2023. [paper] - AutoSTDiff (
xu2025autostdiff) AutoSTDiff: Autoregressive Spatio-Temporal Denoising Diffusion Model for Asynchronous Trajectory Generation. SIAM SDM, 2025. [paper] [code] - DiffMove (
long2025diffmove) DiffMove: Group Mobility Tendency Enhanced Trajectory Recovery via Diffusion Model. arXiv preprint arXiv:2503.18302, 2025. [paper] - GenMove (
long2025one) One Fits All: General Mobility Trajectory Modeling via Masked Conditional Diffusion. arXiv preprint arXiv:2501.13347, 2025. [paper] - Diff-DGMN (
zuo2024diff) Diff-DGMN: A Diffusion-Based Dual Graph Multiattention Network for POI Recommendation. IEEE Internet of Things Journal, 2024. [paper] [code] - DCPR (
long2024diffusion) Diffusion-Based Cloud-Edge-Device Collaborative Learning for Next POI Recommendations. KDD, 2024. [paper] - Traveller (
luo2025traveller) Traveller: Travel-Pattern Aware Trajectory Generation via Autoregressive Diffusion Models. Information Fusion, 2025. [paper] [code] - TrajGDM (
chu2024simulating) Simulating Human Mobility with a Trajectory Generation Framework Based on Diffusion Model. International Journal of Geographical Information Science, 2024. [paper] [code]
-
As Representor
- Poi-enhancer (
cheng2025poi) Poi-enhancer: An LLM-based semantic enhancement framework for POI representation learning. AAAI, 2025. [paper] [code] - LLM-Mob (
wang2023would) Where would i go next? large language models as human mobility predictors. arXiv preprint arXiv:2308.15197, 2023. [paper] [code] - TrajCogn (
zhou2024trajcogn) TrajCogn: Leveraging LLMs for Cognizing Movement Patterns and Travel Purposes from Trajectories. arXiv preprint arXiv:2405.12459, 2024. [paper] [code]
- Poi-enhancer (
-
As Predictor
- TPP-LLM (
liu2024tpp) Tpp-llm: Modeling temporal point processes by efficiently fine-tuning large language models. arXiv preprint arXiv:2410.02062, 2024. [paper] [code] - CoMaPOI (
zhong2025comapoi) CoMaPOI: A Collaborative Multi-Agent Framework for Next POI Prediction Bridging the Gap Between Trajectory and Language. SIGIR, 2025. [paper] [code] - AgentMove (
feng2024agentmove) Agentmove: A large language model based agentic framework for zero-shot next location prediction. NAACL, 2024. [paper] [code] - Feng et al. (
feng2024move) Where to move next: Zero-shot generalization of llms for next poi recommendation. IEEE CAI, 2024. [paper] [code] - LLM4Poi (
li2024large) Large language models for next point-of-interest recommendation. SIGIR, 2024. [paper] [code] - CSA-Rec (
wang2025collaborative) Collaborative Semantics-Assisted Large Language Models for Next POI Recommendation. ICASSP, 2025. [paper] - LAMP (
balsebre2024lamp) LAMP: A language model on the map. arXiv preprint arXiv:2403.09059, 2024. [paper] [code] - Zhang et al. (
zhang2023large) Large Language Models for Spatial Trajectory Patterns Mining.(2023). ACM SIGSPATIAL International Workshop on Geospatial Anomaly Detection, 2024. [paper] - Mo et al. (
mo2023large) Large language models for travel behavior prediction. arXiv preprint arXiv:2312.00819, 2023. [paper] - POI GPT (
kim2024poi) POI GPT: Extracting POI information from social media text data. ISPRS Archives, 2024. [paper] [code] - Chen et al. (
chen2025toward) Toward interactive next location prediction driven by large language models. IEEE Transactions on Computational Social Systems, 2025. [paper] - DelayPTC-LLM (
chen2024delayptc) Delayptc-llm: Metro passenger travel choice prediction under train delays with large language models. arXiv preprint arXiv:2410.00052, 2024. [paper]
- TPP-LLM (
-
Generator
- Liu et al. (
liu2025aligning) Aligning LLM agents with human learning and adjustment behavior: a dual agent approach. arXiv, 2025. [paper] - CoPB (
shao2024chain) Chain-of-planned-behaviour workflow elicits few-shot mobility generation in llms. arXiv, 2024. [paper] [code] - Liu et al. (
liu2023can) Can language models be used for real-world urban-delivery route optimization?. The Innovation, 2023. [paper] - Bhandari et al. (
bhandari2024urban) Urban mobility assessment using llms. SIGSPATIAL, 2024. [paper] - Zheng et al. (
zheng2025urban) Urban planning in the era of large language models. Nature computational science, 2025. [paper]
- Liu et al. (
-
LLM Agents
- LLM-HABG (
meng2025behavior) Behavior Generation for Heterogeneous Agents in Urban Simulation Deduction: A Multi-Stage Approach Based on Large Language Models. CCSSTA, 2025. [paper] - PathGPT (
marcelyn2025pathgpt) PathGPT: Leveraging Large Language Models for Personalized Route Generation. arXiv preprint arXiv:2504.05846, 2025. [paper] [code] - LLMTraveler (
wang2024ai) Ai-driven day-to-day route choice. arXiv preprint arXiv:2412.03338, 2024. [paper] [code] - GATSim (
liu2025gatsim) GATSim: Urban Mobility Simulation with Generative Agents. arXiv preprint arXiv:2506.23306, 2025. [paper] [code] - MobAgent (
li2024more) Be more real: Travel diary generation using llm agents and individual profiles. arXiv preprint arXiv:2407.18932, 2024. [paper] - CitySim (
bougie2025citysim) CitySim: Modeling Urban Behaviors and City Dynamics with Large-Scale LLM-Driven Agent Simulation. EMNLP, 2025. [paper] - TravelPlanner (
xie2024travelplanner) Travelplanner: A benchmark for real-world planning with language agents. arXiv preprint arXiv:2402.01622, 2024. [paper] [code] - IDM-GPT (
yang2025independent) Independent mobility gpt (idm-gpt): A self-supervised multi-agent large language model framework for customized traffic mobility analysis using machine learning models. arXiv preprint arXiv:2502.18652, 2025. [paper]
- LLM-HABG (
-
Encoder-based (BERT-like)
-
Decoder-based (GPT-like)
- MotionLM (
seff2023motionlm) MotionLM: Multi-agent motion forecasting as language modeling. ICCV, 2023. [paper] - RAW (
zhang2023regions) Regions are who walk them: a large pre-trained spatiotemporal model based on human mobility for ubiquitous urban sensing. arXiv preprint arXiv:2311.10471, 2023. [paper] [code]
- MotionLM (
-
Encoder–Decoder-based
- LMTraj (
bae2024can) Can language beat numerical regression? language-based multimodal trajectory prediction. CVPR, 2024. [paper] [[code]](https: //github.com/inhwanbae/LMTrajectory) - RouteLLM (
hallgarten2025routellm) RouteLLM: A Large Language Model with Native Route Context Understanding to Enable Context-Aware Reasoning. IMWUT, 2025. [paper] [code] - QT-Mob (
chen2025enhancing) Enhancing Large Language Models for Mobility Analytics with Semantic Location Tokenization. KDD, 2025. [paper] [code] - CAMS (
du2025cams) CAMS: A CityGPT-Powered Agentic Framework for Urban Human Mobility Simulation. arXiv preprint arXiv:2506.13599, 2025. [paper] - AutoTimes (
liu2024autotimes) AutoTimes: Autoregressive time series forecasters via large language models. NeurIPS, 2024. [paper] [code]
- BERT4Traj (
yang2025bert4traj) BERT4Traj: Transformer Based Trajectory Reconstruction for Sparse Mobility Data. arXiv preprint arXiv:2507.03062, 2025. [paper] - EETG-SVAE (
zhang2025end) End-to-end Trajectory Generation - Contrasting Deep Generative Models and Language Models. ACM Transactions on Spatial Algorithms and Systems, 2025. [paper] - Musleh et al. (
musleh2022towards) Towards a Unified Deep Model for Trajectory Analysis. SIGSPATIAL, 2022. [paper] - UrbanGPT (
li2024urbangpt) UrbanGPT: Spatio-Temporal Large Language Models. KDD, 2024. [paper] [code] - UniST (
yuan2024unist) UniST: A Prompt-Empowered Universal Model for Urban Spatio-Temporal Prediction. KDD, 2024. [paper] [code] - FlashST (
li2024flashst) FlashST: A Simple and Universal Prompt-Tuning Framework for Traffic Prediction. arXiv preprint arXiv:2405.17898, 2024. [paper] [code] - Traffic-Twitter Transformer (
tsai2022traffic) Traffic-Twitter Transformer: A Nature Language Processing-joined Framework for Network-wide Traffic Forecasting. arXiv preprint arXiv:2206.11078, 2022. [paper] - FlowDistill (
yu2025flowdistill) FlowDistill: Scalable Traffic Flow Prediction via Distillation from LLMs. arXiv preprint arXiv:2504.02094, 2025. [paper] [code] - Cao et al. (
cao2021bert) BERT-Based Deep Spatial-Temporal Network for Taxi Demand Prediction. T-ITS, 2021. [paper] - Ma et al. (
ma2025urban) Urban rail transit passenger flow prediction using large language model under multi-source spatiotemporal data fusion. Physica A: Statistical Mechanics and its Applications, 2025. [paper] - TrafficBERT (
jin2021trafficbert) TrafficBERT: Pre-trained model with large-scale data for long-range traffic flow forecasting. Expert Systems with Applications, 2021. [paper] - ST-LLM+ (
liu2025st) ST-LLM+: Graph Enhanced Spatio-Temporal Large Language Models for Traffic Prediction. IEEE Transactions on Knowledge and Data Engineering, 2025. [paper] [code] - MDTI (
liu2025multimodal) Multimodal Trajectory Representation Learning for Travel Time Estimation. arXiv preprint arXiv:2510.05840, 2025. [paper] [code]
- TPLLM (
ren2024tpllm) TPLLM: A traffic prediction framework based on pretrained large language models. arXiv preprint arXiv:2403.02221, 2024. [paper] - LLM-TFP (
cheng2025llm) LLM-TFP: Integrating large language models with spatio-temporal features for urban traffic flow prediction. Applied Soft Computing, 2025. [paper] [code] - Liao et al. (
liao2025next) Next-Generation Travel Demand Modeling with a Generative Framework for Household Activity Coordination. arXiv preprint arXiv:2507.08871, 2025. [paper]
-
Representation and Mining
- Zhang et al. (
zhang2024large) Large language models for spatial trajectory patterns mining. SIGSPATIAL, 2024. [paper] [code] - GPT-J (
ji2024evaluating) Evaluating the Effectiveness of Large Language Models in Representing and Understanding Movement Trajectories. arXiv preprint arXiv:2409.00335, 2024. [paper] - GeoLLM (
manvi2023geollm) Geollm: Extracting geospatial knowledge from large language models. arXiv preprint arXiv:2310.06213, 2023. [paper] [code] - AuxMobLCast (
xue2022leveraging) Leveraging language foundation models for human mobility forecasting. SIGSPATIAL, 2022. [paper] [code] - Wang et al. (
wang2025event) Event-aware analysis of cross-city visitor flows using large language models and social media data. arXiv preprint arXiv:2505.03847, 2025. [paper]
- Zhang et al. (
-
Prediction
- LLM-MPE (
liang2024exploring) Exploring large language models for human mobility prediction under public events. Computers, Environment and Urban Systems, 2024. [paper] - STCInterLLM (
li2025causal) Causal Intervention Is What Large Language Models Need for Spatio-Temporal Forecasting. IEEE Transactions on Cybernetics, 2025. [paper] [code] - xTP-LLM (
guo2024towards) Towards explainable traffic flow prediction with large language models. Communications in Transportation Research, 2024. [paper] [code] - Cai et al. (
cai2024temporal) Temporal-Spatial Traffic Flow Prediction Model Based on Prompt Learning. ISPRS International Journal of Geo-Information, 2025. [paper] - LLM4PT (
wu2025llm4pt) LLM4PT: A large language model-based system for flexible and explainable public transit demand prediction. Computers & Industrial Engineering, 2025. [paper] - TransLLM (
leng2025transllm) TransLLM: A Unified Multi-Task Foundation Framework for Urban Transportation via Learnable Prompting. arXiv preprint arXiv:2508.14782, 2025. [paper] [code]
- LLM-MPE (
-
Generation
- CoDiffMob (
codiffmob2025) Noise Matters: Diffusion Model-Based Urban Mobility Generation with Collaborative Noise Priors. WWW, 2025. [paper] [code] - ControlTraj (
zhu2024controltraj) ControlTraj: Controllable Trajectory Generation with Topology-Constrained Diffusion Model. KDD, 2024. [paper] [code] - DiffTraj (
difftraj2023) DiffTraj: Generating GPS Trajectory with Diffusion Probabilistic Model. NeurIPS, 2023. [paper] [code] - Cardiff (
guo2025leveraging) Leveraging the Spatial Hierarchy: Coarse-to-fine Trajectory Generation via Cascaded Hybrid Diffusion. arXiv preprint arXiv:2507.13366, 2025. [paper] [code] - UniMob (
long2025universal) A Universal Model for Human Mobility Prediction. KDD, 2025. [paper] [code]
- UniFlow (
yuan2024uniflow) UniFlow: A Foundation Model for Unified Urban Spatio-Temporal Flow Prediction. arXiv preprint arXiv:2411.12972, 2024. [paper] [code] - RePST (
wang2024repst) RePST: Language Model Empowered Spatio-Temporal Forecasting via Semantic-Oriented Reprogramming. IJCAI, 2025. [paper] [code] - CompactST (
han2025scalable) Scalable Pre-Training of Compact Urban Spatio-Temporal Predictive Models on Large-Scale Multi-Domain Data. VLDB, 2025. [paper] [code] - STD-PLM (
huang2025std) Std-PLM: Understanding Both Spatial and Temporal Properties of Spatial-Temporal Data with PLM. AAAI, 2025. [paper] [code]
- STG-LLM (
liu2024can) Can Large Language Models Capture Human Travel Behavior? Evidence and Insights on Mode Choice. SSRN, 2024. [paper] - ST-LLM (
liu2024spatial) Spatial-Temporal Large Language Model for Traffic Prediction. MDM, 2024. [paper] [code]
- STGormer (
zhou2024navigating) Navigating Spatio-Temporal Heterogeneity: A Graph Transformer Approach for Traffic Forecasting. arXiv preprint arXiv:2408.10822, 2024. [paper] [code] - STGLLM-E (
rong2024edge) Edge Computing Enabled Large-Scale Traffic Flow Prediction With GPT in Intelligent Autonomous Transport System for 6G Network. T-ITS, 2024. [paper] - CityCAN (
wang2024citycan) CityCAN: Causal Attention Network for Citywide Spatio-Temporal Forecasting. WSDM, 2024. [paper] - STTNs (
xu2020spatial) Spatial-Temporal Transformer Networks for Traffic Flow Forecasting. arXiv preprint arXiv:2001.02908, 2020. [paper] - ST-LINK (
jeon2025st) ST-LINK: Spatially-Aware Large Language Models for Spatio-Temporal Forecasting. CIKM, 2025. [paper] [code]
- UrbanGPT (
li2024urbangpt) UrbanGPT: Spatio-temporal large language models. KDD, 2024. [paper] [code]
- LEAF (
zhao2024embracing) Embracing large language models in traffic flow forecasting. Findings of the Association for Computational Linguistics: ACL 2025, 2025. [paper] [code] - LLMCOD (
yu2024harnessing) Harnessing llms for cross-city od flow prediction. SIGSPATIAL, 2024. [paper] - TraffiCoT-R (
alsahfi2025trafficot) TraffiCoT-R: A framework for advanced spatio-temporal reasoning in large language models. Alexandria Engineering Journal, 2025. [paper]
- DiffODGen (
rong2023complexity) Complexity-aware large scale origin-destination network generation via diffusion model. arXiv preprint arXiv:2306.04873, 2023. [paper] - OpenDiff (
chai2024diffusion) Diffusion model-based mobile traffic generation with open data for network planning and optimization. KDD, 2024. [paper] [code] - Rong et al. (
ronglarge) A Large-scale Dataset and Benchmark for Commuting Origin-Destination Flow Generation. ICLR, 2025. [paper] [code]
- UrbanLLaVA (
feng2025urbanllava) UrbanLLaVA: A Multi-modal Large Language Model for Urban Intelligence with Spatial Reasoning and Understanding. arXiv preprint arXiv:2506.23219, 2025. [paper] [code] - Traj-MLLM (
liu2025traj) Traj-MLLM: Can Multimodal Large Language Models Reform Trajectory Data Mining?. arXiv preprint arXiv:2509.00053, 2025. [paper] - Flame (
xu2025flame) Flame: Learning to Navigate with Multimodal LLM in Urban Environments. AAAI, 2025. [paper] [code] - VLMLocPredictor (
zhang2025eyes) Eyes Will Shut: A Vision-Based Next GPS Location Prediction Model by Reinforcement Learning from Visual Map Feed Back. arXiv preprint arXiv:2507.18661, 2025. [paper] [code] - MapGPT (
chen2024mapgpt) MapGPT: Map-guided Prompting with Adaptive Path Planning for Vision-and-Language Navigation. Annual Meeting of the Association for Computational Linguistics, 2024. [paper] [code] - UGI (
xu2023urban) Urban Generative Intelligence (UGI): A Foundational Platform for Agents in Embodied City Environment. arXiv preprint arXiv:2312.11813, 2023. [paper] [code] - CityBench (
feng2025citybench) CityBench: Evaluating the Capabilities of Large Language Models as World Models for Urban Tasks. KDD, 2025. [paper] [code] - LLM-enhanced POI recommendation (
wang2025beyond) Beyond Visit Trajectories: Enhancing POI Recommendation via LLM-Augmented Text and Image Representations. RecSys, 2025. [paper] [code]
- TrajSceneLLM (
ji2025trajscenellm) TrajSceneLLM: A Multimodal Perspective on Semantic GPS Trajectory Analysis. arXiv preprint arXiv:2506.16401, 2025. [paper] [code] - Path-LLM (
wei2025path) Path-LLM: A Multi-Modal Path Representation Learning by Aligning and Fusing with Large Language Models. WWW, 2025. [paper] [code] - Trajectory-LLM (
yang2025trajectory) Trajectory-LLM: A Language-Based Data Generator for Trajectory Prediction in Autonomous Driving. ICLR, 2025. [paper] [code] - TrajAgent (
du2024trajagent) TrajAgent: An LLM-based Agent Framework for Automated Trajectory Modeling via Collaboration of Large and Small Models. arXiv preprint arXiv:2410.20445, 2024. [paper] [code] - CoAST (
zhai2025cognitive) Cognitive-Aligned Spatio-Temporal Large Language Models For Next Point-of-Interest Prediction. arXiv preprint arXiv:2510.14702, 2025. [paper] - CityGPT (
feng2025citygpt) CityGPT: Empowering Urban Spatial Cognition of Large Language Models. KDD, 2025. [paper] [code] - POI-Enhancer (
cheng2025poi) POI-Enhancer: An LLM-based Semantic Enhancement Framework for POI Representation Learning. AAAI, 2025. [paper] [code] - D2A (
wang2024simulating) Simulating Human-like Daily Activities with Desire-driven Autonomy. arXiv preprint arXiv:2412.06435, 2024. [paper] [code]
- Vision-LLM (
yang2025vision) Vision-LLMs for Spatiotemporal Traffic Forecasting. arXiv preprint arXiv:2510.11282, 2025. [paper] - OpenDiff (
chai2024diffusion) Diffusion Model-based Mobile Traffic Generation with Open Data for Network Planning and Optimization. KDD, 2024. [paper] [code] - LSDM (
zhang2025lsdm) LSDM: LLM-Enhanced Spatio-temporal Diffusion Model for Service-Level Mobile Traffic Prediction. arXiv preprint arXiv:2507.17795, 2025. [paper] [code]
- ChatTraffic (
zhang2024chattraffic) ChatTraffic: Text-to-Traffic Generation via Diffusion Model. T-ITS, 2024. [paper] [code] - ChatSUMO (
li2024chatsumo) ChatSUMO: Large Language Model for Automating Traffic Scenario Generation in Simulation of Urban Mobility. IEEE Transactions on Intelligent Vehicles, 2024. [paper] - UrbanMind (
liu2025urbanmind) UrbanMind: Urban Dynamics Prediction with Multifaceted Spatial-Temporal Large Language Models. KDD, 2025. [paper] [code] - T3 (
han2024event) Event Traffic Forecasting with Sparse Multimodal Data. ACM MM, 2024. [paper] [code] - GPT4MTS (
jia2024gpt4mts) GPT4MTS: Prompt-based Large Language Model for Multimodal Time-Series Forecasting. AAAI, 2024. [paper]
- Sat2Flow (
wang2025sat2flow) Sat2Flow: A Structure-Aware Diffusion Framework for Human Flow Generation from Satellite Imagery. arXiv preprint arXiv:2508.19499, 2025. [paper] [code] - GlODGen (
rong2025satellites) Satellites Reveal Mobility: A Commuting Origin-Destination Flow Generator for Global Cities. arXiv preprint arXiv:2505.15870, 2025. [paper] [code]
- SeMob (
chen2025semob) SeMob: Semantic Synthesis for Dynamic Urban Mobility Prediction. EMNLP, 2025. [paper] [code] - Ernie-GeoL (
huang2022ernie) ERNIE-GeoL: A Geography-and-Language Pre-trained Model and Its Applications in Baidu Maps. KDD, 2022. [paper] - FUSE-Traffic (
yu2025fuse) FUSE-Traffic: Fusion of Unstructured and Structured Data for Event-aware Traffic Forecasting. ACM SIGSPATIAL, 2025. [paper] [code] - CityFM (
balsebre2024city) City Foundation Models for Learning General Purpose Representations from OpenStreetMap. CIKM, 2024. [paper] [code]
| Category | Dataset | Description | Geography | Statistics | Year |
|---|---|---|---|---|---|
| Discrete Sequence | Veraset-Visits | Individual-level POI check-ins | USA | 4+ million points of interest | 2019-Present |
| Discrete Sequence | Taxi Trips [Link1][Link2] | Taxi trip records | Chicago | 224.8 millions trips | 2013-Present |
| Discrete Sequence | NYC TLC | Taxi trip records | NYC | Billions of trips | 2009-2025 |
| Discrete Sequence | Tencent Mobility[Link1][Link2] | POI check-ins | Beijing | 297,363,263 trajectory points | 2019 |
| Discrete Sequence | ChinaMobile[Link1][Link2] | Cellular trajectories | Beijing | 4,163,651 points from 1,246 users | 2017 |
| Discrete Sequence | GMove | Tweet check-in trajectories | New York, Los Angeles | 72 thousand trajectories | 2014 |
| Discrete Sequence | Foursquare-Global [Link1][Link2] | Individual-level POI check-ins | Global | 33,278,683 check-ins | 2012-2013 |
| Discrete Sequence | Brightkite | Individual-level POI check-ins | Global | 4,491,143 check-ins | 2008-2010 |
| Discrete Sequence | Gowalla | Individual-level POI check-ins | Global | 6,442,890 checkins | 2009-2010 |
| Discrete Sequence | Yelp-check-ins | POI-level check-ins | 11 metropolitan areas | 150,346 businesses | - |
| Continuous Sequence | MTA Subway Ridership [Link1][Link2] | Subway ridership | NYC | 178.6 millions Records | 2017-present |
| Continuous Sequence | Advan Weekly Patterns | POI-level aggregated temporal visit metrics | USA, Canada | updated weekly | 2018-Present |
| Continuous Sequence | Advan Monthly Patterns | POI-level aggregated temporal visit metrics | USA | updated monthly | 2019-Present |
| Continuous Sequence | DiDi Chuxing | Ride-hailing trajectories | Xi'an, Chengdu | 8 billion trajectories | 2016 |
| Continuous Sequence | TaxiPorto [Link1][Link2][Link3] | GPS taxi trajectories | Porto | 442 taxis | 2013-2014 |
| Continuous Sequence | GeoLife | Human GPS trajectories | Cities in China, USA, and Europe | 17,621 trajectories | 2007-2012 |
| Continuous Sequence | T-Drive [Link1][Link2] | Taxi GPS trajectories | Beijing | 169,984 trajectories, 10,357 taxis | 2008 |
| Continuous Sequence | YJMob100K | Human trajectories | Japan | 111,535,175 or 29,389,749 records | - |
| Spatio-Temporal Graph | PEMS [Link1] [Link2] | Mobility network | California | 2 GB per day, 39,000+ detectors | 1998-Present |
| Spatio-Temporal Graph | NYC Yellow Taxi [Link1][Link2] | Hourly trip count tensors | NYC | 36.4 million trip volumes | 2011-2024 |
| Spatio-Temporal Graph | CHI-Taxi [Link1][Link2] | Taxi OD flow network | Chicago | 77 nodes | 2023 |
| Spatio-Temporal Graph | UK mobility flow | Commuting OD flow network | UK | - | 1981,1991,2001,2011,2021 |
| Spatio-Temporal Graph | Italy mobility flow | Commuting OD flow network | Italy | - | 1991,2001,2011,2021 |
| Spatio-Temporal Graph | LargeST | Vehicle traffic flows | California | 525,888 time frames | 2017-2021 |
| Spatio-Temporal Graph | COVID-19 Human Mobility | OD flow network | USA | 3 geographic scales | 2019-2021 |
| Spatio-Temporal Graph | LODES-7.5 | Commuting OD flow network | USA | 12 OD files for a state-year | 2002-2019 |
| Spatio-Temporal Graph | TaxiBJ | Taxi flow network | Beijing | 22,459 time intervals | 2013-2016 |
| Spatio-Temporal Graph | BikeNYC [Link1][Link2] | Bike flow network | NYC | Millions Monthly | 2014 |
| Spatio-Temporal Graph | BJER4 [Link1][Link2] | Road network traffic speed dataset | Beijing | 12 roads | 2014 |
| Multimodal Mobility | GlODGen | Vision + OD flows | Global | synthetic data | 2025 |
| Multimodal Mobility | Earth AI | Vision + population + environment | Global | 10-meter resolution | 2025 |
| Multimodal Mobility | BostonWalks | Population + trips + activities | Boston metropolitan area | 155,000 trips, 990 participants | 2023 |
| Multimodal Mobility | TartanAviation | Images + trajectories + speech | Greater Pittsburgh area | 3.1M images, 661 days of trajectories | 2020-2023 |
| Multimodal Mobility | NetMob25 | Population + trip descriptions + trajectories | Greater Paris area | 500 million high frequency points | 2022-2023 |
| Multimodal Mobility | nuScenes | Vision + trajectories | Boston, Singapore | 1000 scenes, 1.4 million images | 2019 |
| Multimodal Mobility | Breadcrumbs | Population + trajectories + POI labeling | Lausanne | 46,380,042 records, 81 users | 2018 |
| Multimodal Mobility | RECORD MultiSensor Study | Population + trajectories + semantic trip annotations | Paris region | 21,163 segments of observation | 2013-2015 |
| Multimodal Mobility | METR-LA [Link1][Link2] | Traffic speed/flows + vehicle traces + event records | Los Angeles | 9,300 sensors covering 5,400 miles | - |
| Multimodal Mobility | LaDe | Delivery & pickup records + road network | Five cities in China | 10,677K packages, 21K couriers | - |
| Multimodal Mobility | Yelp Dataset | Check-ins + text reviews | 11 metropolitan areas | 6,990,280 reviews | - |
| Multimodal Mobility | Waymo Open Motion Dataset | Vision + trajectories | Six U.S. cities | 570+ hours at 10 Hz | - |
@article{Guo_2025,
title={Language Models Meet Urban Mobility: A Data-Centric Review},
url={[http://dx.doi.org/10.36227/techrxiv.176703984.41856875/v1](http://dx.doi.org/10.36227/techrxiv.176703984.41856875/v1)},
DOI={10.36227/techrxiv.176703984.41856875/v1},
publisher={Institute of Electrical and Electronics Engineers (IEEE)},
author={Guo, Baoshen and Hong, Zhiqing and Cao, Lidan and Li, Donghang and Li, Junyi and Rong, Can and Prakash, Alok and Wang, Shenhao and Zhao, Jinhua},
year={2025},
month=dec
}