ICML 2026 | 时空数据（Spatial-Temporal）论文总结（上）【时空预测，信号灯控制，地理空间表示等】

时空探索之旅

发布于 2026-05-18 12:27:00

1230

ICML 2026将在2026年7月6日—11日于韩国首尔（Seoul, South Korea）举行。本文总结了2026 ICML上有关时空数据（Spatial-Temporal）相关论文。如有疏漏，欢迎大家补充。

注：笔者将分为上下2篇推文来总结，本文主要涉及时空数据中有关交通，城市以及地理信息的论文。

本文时空数据Topic：时空（交通）预测，信号灯控制，地理空间表示，城市感知，轨迹预测，LLM/VLM以及MLLM等的应用等。

1. CausalX: A Unified and Causally-Interpretable Plug-and-Play Model for Multi-modal Spatio-Temporal Forecasting2. Decoupling Universal Laws and Environmental Heterogeneity: A Physics-Inspired Framework for Robust Spatio-Temporal Forecasting3. LagLLM: LLM-empowered lead–lag dependency learning for spatial-temporal time series forecasting4. Lightweight and Interpretable Transformer via Unrolling of Mixed Graph Algorithms for Traffic Forecast5. Nested Spatio-Temporal Time Series Forecasting6. Being More Lightweight and Practical: Mini-sized Contrastive Learning Pre-trained Models for Fine-grained Traffic Task7. Spatiotemporal Imputation with Graph-Informed Flow Matching8. Learning Long Range Spatio-Temporal Representations over Continuous Time Dynamic Graphs with State Space Models9. Textual Supervision Enhances Geospatial Representations in Vision-Language Models10. UrbanFusion: Stochastic Multimodal Fusion for Contrastive Learning of Robust Spatial Representations11. Position: Benchmarks for Vision–Language Models in Urban Perception Should Be Reliability-Aware and Negotiated12. UrbanMLLM: Joint Learning of Cross-view Imagery for Urban Understanding13. RAST-MoE-RL: A Regime-Aware Spatio-Temporal MoE Framework for Deep Reinforcement Learning in Ride-Hailing14. Scalable Traffic Signal Control with Shared Policy Framework15. Unlocking Zero-Shot Geospatial Reasoning via Indirect Rewards16. GenDA: Generative Data Assimilation on Complex Urban Areas via Classifier-Free Diffusion Guidance17. TF-FACE: Time-Frequency Fusion Learning via Frequency-Domain Adaptive and Controllable Enhancement for Trajectory Prediction

点击文末阅读原文跳转笔者知乎链接（跳转论文链接更方便）

1 CausalX: A Unified and Causally-Interpretable Plug-and-Play Model for Multi-modal Spatio-Temporal Forecasting

链接：https://icml.cc/virtual/2026/poster/61373

作者：Shiqi Zhang ⋅ Pan Mu ⋅ HantingYan ⋅ Yuchao Zhu ⋅ jinglin zhang ⋅ Cong Bai

关键词：多模态时空预测，因果，可解释性

2 Decoupling Universal Laws and Environmental Heterogeneity: A Physics-Inspired Framework for Robust Spatio-Temporal Forecasting

链接：https://icml.cc/virtual/2026/poster/61390

作者：Aoyu Liu ⋅ Liming Wei ⋅ YAYING ZHANG

关键词：稳健时空预测，物理启发，异质性

3 LagLLM: LLM-empowered lead–lag dependency learning for spatial-temporal time series forecasting

链接：https://icml.cc/virtual/2026/poster/66176

作者：Binqing Wu ⋅ Jian Zhou ⋅ Zongjiang Shang ⋅ Ling Chen

关键词：时空预测，LLM，滞后依赖

4 Lightweight and Interpretable Transformer via Unrolling of Mixed Graph Algorithms for Traffic Forecast

链接：https://icml.cc/virtual/2026/poster/63731

作者：Ji Qi ⋅ Mingxiao Liu ⋅ VIET THUC ⋅ Yuzhe Li ⋅ Zhuoshi Pan ⋅ Gene Cheung ⋅ Hong Zhao

关键词：时空预测，轻量化，可解释性，Transformer

5 Nested Spatio-Temporal Time Series Forecasting

链接：https://icml.cc/virtual/2026/poster/60735

作者：YingHao Ai ⋅ Yukai Zhou ⋅ Ruoxi Jiang ⋅ Junyi An ⋅ chao qu ⋅ Zhijian Zhou ⋅ Shiyu Wang ⋅ Fenglei Cao ⋅ Zenglin Xu ⋅ Furao Shen ⋅ Yuan Qi

关键词：嵌套时空预测

6 Being More Lightweight and Practical: Mini-sized Contrastive Learning Pre-trained Models for Fine-grained Traffic Task

链接：https://icml.cc/virtual/2026/poster/66525

作者：Shuhao Li ⋅ Weidong Yang ⋅ Ben Fei ⋅ Yue Cui ⋅ Lipeng Ma ⋅ Fan Zhang

关键词：细粒度交通预测

7 Spatiotemporal Imputation with Graph-Informed Flow Matching

链接：https://icml.cc/virtual/2026/poster/65534

作者：Zepeng Zhang ⋅ Aref Einizade ⋅ Jhony Giraldo ⋅ Olga Fink

关键词：插补，流匹配

8 Learning Long Range Spatio-Temporal Representations over Continuous Time Dynamic Graphs with State Space Models

链接：https://icml.cc/virtual/2026/poster/63979

作者：Ayushman Raghuvanshi ⋅ Thummaluru Reddy ⋅ Sundeep Prabhakar Chepuri ⋅ Mahesh Chandran

关键词：时空表示，状态空间模型，动态图

9 Textual Supervision Enhances Geospatial Representations in Vision-Language Models

链接：https://icml.cc/virtual/2026/poster/60539

作者：Marcelo Sartori Locatelli ⋅ Fernando Tonucci ⋅ Jea Kwon ⋅ Luiz Felipe Vecchietti ⋅ Bryan Nathanael Wijaya ⋅ Cheng Yaw Low ⋅ Virgilio Almeida ⋅ MEEYOUNG CHA

关键词：地理空间理解，VLM

10 UrbanFusion: Stochastic Multimodal Fusion for Contrastive Learning of Robust Spatial Representations

链接：https://icml.cc/virtual/2026/poster/65596

作者：Dominik J. Mühlematter ⋅ Lin Che ⋅ Ye Hong ⋅ Martin Raubal ⋅ Nina Wiedemann

关键词：地理空间表示学习，多模态融合，稳健性，对比学习

11 Position: Benchmarks for Vision–Language Models in Urban Perception Should Be Reliability-Aware and Negotiated

链接：https://icml.cc/virtual/2026/poster/67079

作者：Rashid Mushkani

关键词：街景图像生成，VLM

12 UrbanMLLM: Joint Learning of Cross-view Imagery for Urban Understanding

链接：https://icml.cc/virtual/2026/poster/65712

作者：Xin Zhang ⋅ Tianjian Ouyang ⋅ Yu Shang ⋅ Qingmin Liao ⋅ Yong Li

关键词：城市理解，多模态大模型

13 RAST-MoE-RL: A Regime-Aware Spatio-Temporal MoE Framework for Deep Reinforcement Learning in Ride-Hailing

链接：https://icml.cc/virtual/2026/poster/63114

作者：Yuhan Tang ⋅ Kangxin Cui ⋅ Jung Ho Park ⋅ Yibo Zhao ⋅ Xuan Jiang ⋅ Haoze He ⋅ Jiangbo Yu ⋅ Haris Koutsopoulos ⋅ Jinhua Zhao

关键词：MoE，供需匹配

14 Scalable Traffic Signal Control with Shared Policy Framework

链接：https://icml.cc/virtual/2026/poster/63318

作者：Haolun MA ⋅ Yanchen ZHU ⋅ Zizhuo Xu ⋅ Weijie Shi ⋅ Jiajie Xu ⋅ Lei Li

关键词：信控优化，策略感知

15 Unlocking Zero-Shot Geospatial Reasoning via Indirect Rewards

链接：https://icml.cc/virtual/2026/poster/64548

作者：Chenhui Xu ⋅ Fuxun Yu ⋅ Michael Bianco ⋅ Jacob Kovarskiy ⋅ Raphael Tang ⋅ Qi Zhang ⋅ Zirui Xu ⋅ Will LeVine ⋅ Brandon Dubbs ⋅ Heming Liao ⋅ Cassandra Burgess ⋅ Suvam Bag ⋅ Jay Patravali ⋅ Rupanjali Kukal ⋅ Mikael Figueroa ⋅ Rishi Madhok ⋅ Nikolaos Karianakis ⋅ Jinjun Xiong

关键词：地理空间推理，零样本

16 GenDA: Generative Data Assimilation on Complex Urban Areas via Classifier-Free Diffusion Guidance

链接：https://icml.cc/virtual/2026/poster/66132

arXiv：http://arxiv.org/abs/2601.11440v2

作者：Francisco Giral ⋅ Álvaro Sevillano ⋅ Ignacio Perez ⋅ Ricardo Vinuesa ⋅ Soledad Le Clainche

关键词：生成式数据同化，风场重建

17 TF-FACE: Time-Frequency Fusion Learning via Frequency-Domain Adaptive and Controllable Enhancement for Trajectory Prediction

链接：https://icml.cc/virtual/2026/poster/63437

作者：Dongjian Song ⋅ Yunhao Meng ⋅ Songjun Huang ⋅ Jiayi Han

关键词：轨迹预测，频域

如果觉得有帮助还请分享，在看，点赞

本文参与腾讯云自媒体同步曝光计划，分享自微信公众号。

原始发表：2026-05-14，如有侵权请联系 cloudcommunity@tencent.com 删除

virtual

本文分享自时空探索之旅微信公众号，前往查看

如有侵权，请联系 cloudcommunity@tencent.com 删除。

本文参与腾讯云自媒体同步曝光计划，欢迎热爱写作的你一起参与！

登录后参与评论

0 条评论

热度