ICML 2026 | 时间序列（Time Series）论文总结（2）【基础模型，生成，分类，异常检测，插补，表示学习和分析等】

时空探索之旅

发布于 2026-05-18 12:25:16

2190

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上有关时间序列（time series）相关论文。如有疏漏，欢迎大家补充。

注：由于时间序列（标题包含time series或time-series）的论文高达125篇（其中两篇可以算作时空，除去还有123篇），笔者将分为上中下3篇推文来总结，此为第2篇，本文主要涉及时间序列基础模型，生成，分类，异常检测，插补，表示学习和分析等，共计31篇。

本文时间序列Topic：时间序列基础模型（TSFM），生成，分类，异常检测，插补，表示学习和分析等。

1. Olivia: Harmonizing Time Series Foundation Models with Power Spectral Density2. Time-PEFT: Temporal and Multichannel Complexity-Based Fine-Tuning for Time-Series Foundation Models3. Universal Redundancies in Time Series Foundation Models4. Time Series Reasoning via Process-Verifiable Thinking Data Synthesis and Scheduling for Tailored LLM Reasoning5. OpenTSLM: Time-Series Language Models for Reasoning over Multivariate Medical Text- and Time-Series Data6. HEARTS: Benchmarking LLM Reasoning on Health Time Series7. Position: Time-Series Foundation Models Require Explicit Domain-Level Benchmarks8. StarEmbed: Benchmarking Time Series Foundation Models on Astronomical Observations of Variable Stars9. Winformer: Transcending Pairwise Similarity for Time-series Generation10. ConTSG-Bench: A Unified Benchmark for Conditional Time Series Generation11. MN-Diff: Diffusion Parameterized MoE-NCDE for Continuous Time Series Generation with Irregular Observations12. TimeOmni-VL: Unified Models for Time Series Understanding and Generation13. CURE: Context-driven Diffusion with Progressive Expansion for Single Domain Generalization in Time Series Classification14. Time-CoT: Hierarchical Reasoning with Temporal Semantic Codes for Multivariate Time Series Classification15. MedMamba: Multi-View State Space Models with Adaptive Graph Learning for Medical Time Series Classification16. Mantis: Lightweight Foundation Model for Time Series Classification17. One-Step Graph-Structured Neural Flows for Irregular Multivariate Time Series Classification18. Cluster-Aware Causal Mixer for Online Anomaly Detection in Multivariate Time Series19. COGNOS: Universal Enhancement for Time Series Anomaly Detection via Constrained Gaussian-Noise Optimization and Smoothing20. AnomSeer: Reinforcing Multimodal LLMs to Reason for Time-Series Anomaly Detection21. IMPACT: Influence Modeling for Open-Set Time Series Anomaly Detection22. Towards Foundation Models for Zero-Shot Time Series Anomaly Detection: Leveraging Synthetic Data and Relative Context Discrepancy23. Online Change Point Detection for Multivariate Inhomogeneous Poisson Processes Time Series24. TeamWork: Multivariate Time Series Anomaly Detection via Asymmetric Role-aware Channel Modeling25. HELIX: Hybrid Encoding with LearnableIdentityand Cross-dimensional Synthesis for Time Series Imputation26. Rethinking Time-Series Imputation as Conditional Inference along Temporal Evolution27. Self-Supervised Dynamical System Representations for Physiological Time-Series28. One Batch Is Enough: A Unified Dataset Condensation Framework for General Time Series Analysis29. Zeus: Towards Tuning-Free Foundation Model for Time Series Analysis30. TelecomTS: A Multi-Modal Observability Dataset for Time Series and Language Analysis31. Time Series, Vision, and Language: Exploring the Limits of Alignment in Contrastive Representation Spaces

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 Olivia: Harmonizing Time Series Foundation Models with Power Spectral Density

链接：https://icml.cc/virtual/2026/poster/65289

作者：Jingru Fei ⋅ Kun Yi ⋅ Alex Wang ⋅ Qingsong Wen ⋅ Xiangxiang Zhu ⋅ Wei Fan

关键词：基础模型，对齐融合，频谱

2 Time-PEFT: Temporal and Multichannel Complexity-Based Fine-Tuning for Time-Series Foundation Models

链接：https://icml.cc/virtual/2026/poster/61767

作者：Jihye Na ⋅ Patara Trirat ⋅ Chanyoung Park ⋅ Jae-Gil Lee

关键词：基础模型，参数高效微调

3 Universal Redundancies in Time Series Foundation Models

链接：https://icml.cc/virtual/2026/poster/65406

arXiv：http://arxiv.org/abs/2602.01605v1

代码：https://github.com/abao1999/tsfm-lens

作者：Anthony Bao ⋅ Venkata Hasith Vattikuti ⋅ Jeffrey Lai ⋅ William Gilpin

关键词：基础模型，可解释性，压缩

4 Time Series Reasoning via Process-Verifiable Thinking Data Synthesis and Scheduling for Tailored LLM Reasoning

链接：https://icml.cc/virtual/2026/poster/64395

arXiv：http://arxiv.org/abs/2602.07830v2

作者：Jiahui Zhou ⋅ Dan Li ⋅ Boxin Li ⋅ Xiao Zhang ⋅ Erli Meng ⋅ Lin Li ⋅ Zhuomin Chen ⋅ Jian Lou ⋅ See-Kiong Ng

关键词：时序推理

5 OpenTSLM: Time-Series Language Models for Reasoning over Multivariate Medical Text- and Time-Series Data

链接：https://icml.cc/virtual/2026/poster/65261

arXiv：http://arxiv.org/abs/2510.02410v3

作者：Patrick Langer ⋅ Thomas Kaar ⋅ Max Rosenblattl ⋅ Maxwell Xu ⋅ Winnie Chow ⋅ Martin Maritsch ⋅ Robert Jakob ⋅ Ning Wang ⋅ Juncheng Liu ⋅ Aradhana Verma ⋅ Brian Han ⋅ Daniel Kim ⋅ Henry Chubb ⋅ Scott Ceresnak ⋅ Aydin Zahedivash ⋅ Alexander Sandhu ⋅ Fatima Rodriguez ⋅ Daniel McDuff ⋅ Elgar Fleisch ⋅ Oliver Aalami ⋅ Filipe Barata ⋅ Paul Schmiedmayer

关键词：时序大模型，医疗文本时序

6 HEARTS: Benchmarking LLM Reasoning on Health Time Series

链接：https://icml.cc/virtual/2026/poster/61389

arXiv：http://arxiv.org/abs/2603.06638v2

代码：https://github.com/yang-ai-lab/HEARTS

作者：Sirui Li ⋅ Shuhan Xiao ⋅ Mihir Joshi ⋅ Ahmed Metwally ⋅ Daniel McDuff ⋅ Wei Wang ⋅ Yuzhe Yang

关键词：benchmark，健康时序，时序推理

7 Position: Time-Series Foundation Models Require Explicit Domain-Level Benchmarks

链接：https://icml.cc/virtual/2026/poster/67138

作者：Asif Bin Syed ⋅ Md Younus Ahamed ⋅ Azmine Toushik Wasi

关键词：领域级基准

8 StarEmbed: Benchmarking Time Series Foundation Models on Astronomical Observations of Variable Stars

链接：https://icml.cc/virtual/2026/poster/63677

arXiv：http://arxiv.org/abs/2510.06200v3

代码：https://github.com/skai-institute/StarEmbed

作者：Weijian Li ⋅ Hong-Yu Chen ⋅ Nabeel Rehemtulla ⋅ Ved Shah ⋅ Dongho Kim ⋅ Dennis Wu ⋅ Qinjie Lin ⋅ Adam Miller ⋅ Han Liu

关键词：benchmark，恒星时间序列观测

9 Winformer: Transcending Pairwise Similarity for Time-series Generation

链接：https://icml.cc/virtual/2026/poster/63196

作者：Haoyi Zhou ⋅ Xin Xue ⋅ Tianyu Chen ⋅ lanhao li ⋅ Lijun SUN ⋅ Jianxin Li

关键词：生成，周期错位

10 ConTSG-Bench: A Unified Benchmark for Conditional Time Series Generation

链接：https://icml.cc/virtual/2026/poster/64506

arXiv：http://arxiv.org/abs/2603.04767v1

代码：https://github.com/seqml/ConTSG-Bench

作者：Shaocheng Lan ⋅ Shuqi Gu ⋅ Zhangzhi Xiong ⋅ Kan Ren

关键词：生成，周期错位

11 MN-Diff: Diffusion Parameterized MoE-NCDE for Continuous Time Series Generation with Irregular Observations

链接：https://icml.cc/virtual/2026/poster/61786

作者：Xu Zhang ⋅ Junwei Deng ⋅ Chang Xu ⋅ Hao Li ⋅ Jiang Bian

关键词：（持续）生成，MoE，不规则观察

12 TimeOmni-VL: Unified Models for Time Series Understanding and Generation

链接：https://icml.cc/virtual/2026/poster/61019

arXiv：http://arxiv.org/abs/2602.17149v1

作者：Tong Guan ⋅ SHENG PAN ⋅ Johan Barthelemy ⋅ Zhao Li ⋅ Yujun Cai ⋅ Cesare Alippi ⋅ Ming Jin ⋅ Shirui Pan

关键词：生成，理解，视觉模态

13 CURE: Context-driven Diffusion with Progressive Expansion for Single Domain Generalization in Time Series Classification

链接：https://icml.cc/virtual/2026/poster/65631

作者：Yuhang Pei ⋅ Fanchun Meng ⋅ Wenrui Wu ⋅ Tao Ren ⋅ Yifan Wang ⋅ Wei Ju ⋅ Chao Zheng ⋅ Xiao Luo

关键词：分类，扩散模型

14 Time-CoT: Hierarchical Reasoning with Temporal Semantic Codes for Multivariate Time Series Classification

链接：https://icml.cc/virtual/2026/poster/62390

作者：Kun Zeng ⋅ Wu Binquan ⋅ Qianli Ma

关键词：分类，时序思维链

15 MedMamba: Multi-View State Space Models with Adaptive Graph Learning for Medical Time Series Classification

链接：https://icml.cc/virtual/2026/poster/61414

作者：Da Zhang ⋅ bingyu li ⋅ Zhiyuan Zhao ⋅ Hongyuan Zhang ⋅ Junyu Gao ⋅ Xuelong Li

关键词：医疗时序分类，Mamba，自适应图学习

16 Mantis: Lightweight Foundation Model for Time Series Classification

链接：https://icml.cc/virtual/2026/poster/62438

作者：Vasilii Feofanov ⋅ Songkang Wen ⋅ Shifeng Xie ⋅ Simon Roschmann ⋅ Marius Alonso ⋅ Hongbo Guo ⋅ Romain Ilbert ⋅ Malik TIOMOKO ⋅ Quentin Bouniot ⋅ Zeynep Akata ⋅ Lujia Pan ⋅ Jianfeng Zhang ⋅ Ievgen Redko

关键词：分类，轻量化基础模型

17 One-Step Graph-Structured Neural Flows for Irregular Multivariate Time Series Classification

链接：https://icml.cc/virtual/2026/poster/64769

作者：Mengzhou Gao ⋅ kaiwei wang ⋅ Pengfei Jiao

关键词：不规则时序分类，神经流，图结构

18 Cluster-Aware Causal Mixer for Online Anomaly Detection in Multivariate Time Series

链接：https://icml.cc/virtual/2026/poster/64081

arXiv：http://arxiv.org/abs/2506.00188v1

作者：Md Mahmuddun Nabi Murad ⋅ Yasin Yilmaz

关键词：在线异常检测

19 COGNOS: Universal Enhancement for Time Series Anomaly Detection via Constrained Gaussian-Noise Optimization and Smoothing

链接：https://icml.cc/virtual/2026/poster/60841

arXiv：http://arxiv.org/abs/2511.06894v2

作者：Wenlong Shang ⋅ Shihao Tian ⋅ Xutong Wan ⋅ Peng Chang

关键词：异常检测，高斯噪声平滑

20 AnomSeer: Reinforcing Multimodal LLMs to Reason for Time-Series Anomaly Detection

链接：https://icml.cc/virtual/2026/poster/64969

arXiv：http://arxiv.org/abs/2602.08868v1

作者：Junru Zhang ⋅ Lang Feng ⋅ Haoran Shi ⋅ Xu Guo ⋅ Han Yu ⋅ Yabo Dong ⋅ Duanqing Xu

关键词：异常检测，多模态大模型

21 IMPACT: Influence Modeling for Open-Set Time Series Anomaly Detection

链接：https://icml.cc/virtual/2026/poster/62939

作者：Xiaohui Zhou ⋅ Yijie Wang ⋅ Hongzuo Xu ⋅ Weixuan Liang ⋅ Xiaoli Li ⋅ Guansong Pang

关键词：开放集异常检测

22 Towards Foundation Models for Zero-Shot Time Series Anomaly Detection: Leveraging Synthetic Data and Relative Context Discrepancy

链接：https://icml.cc/virtual/2026/poster/60604

arXiv：http://arxiv.org/abs/2509.21190v3

作者：Tian Lan ⋅ Hao Le ⋅ Jinbo Li ⋅ Wenjun He ⋅ Meng Wang ⋅ Chenghao Liu ⋅ Chen Zhang

关键词：异常检测，零样本，基础模型

23 Online Change Point Detection for Multivariate Inhomogeneous Poisson Processes Time Series

链接：https://icml.cc/virtual/2026/poster/64844

arXiv：http://arxiv.org/abs/2601.20192v1

作者：Xiaokai Luo ⋅ Haotian Xu ⋅ Carlos Misael Madrid Padilla ⋅ OSCAR HERNAN MADRID PADILLA

关键词：在线变点检测，多元非齐次泊松点过程

24 TeamWork: Multivariate Time Series Anomaly Detection via Asymmetric Role-aware Channel Modeling

链接：https://icml.cc/virtual/2026/poster/61471

作者：Shiyan Hu ⋅ Tengxue Zhang ⋅ Jianxin Jin ⋅ Xiangfei Qiu ⋅ Bin Yang ⋅ Chenjuan Guo

关键词：异常检测，通道建模

25 HELIX: Hybrid Encoding with LearnableIdentityand Cross-dimensional Synthesis for Time Series Imputation

链接：https://icml.cc/virtual/2026/poster/65249

arXiv：http://arxiv.org/abs/2605.02278v1

代码：https://github.com/milaogou/HELIX

作者：Fengming Zhang ⋅ Wenjie Du ⋅ Huan Zhang ⋅ Ke Yu ⋅ Shen Qu

关键词：插补，跨维度同步，混合编码

26 Rethinking Time-Series Imputation as Conditional Inference along Temporal Evolution

链接：https://icml.cc/virtual/2026/poster/66294

作者：Yu Fan ⋅ Yang Yang ⋅ Yufan Guo ⋅ Huazhong Yang ⋅ pengjun wang

关键词：插补，条件时序推理

27 Self-Supervised Dynamical System Representations for Physiological Time-Series

链接：https://icml.cc/virtual/2026/poster/62877

arXiv：http://arxiv.org/abs/2512.00239v1

作者：Yenho Chen ⋅ Maxwell Xu ⋅ James Rehg ⋅ Christopher Rozell

关键词：生理时序，自监督

28 One Batch Is Enough: A Unified Dataset Condensation Framework for General Time Series Analysis

链接：https://icml.cc/virtual/2026/poster/62269

作者：Wei Shao ⋅ Ziquan Fang ⋅ Zheqi Lu ⋅ Yongfeng Su ⋅ Yuzhu Wang ⋅ Yunjun Gao

关键词：时序分析，数据集浓缩

29 Zeus: Towards Tuning-Free Foundation Model for Time Series Analysis

链接：https://icml.cc/virtual/2026/poster/65415

作者：Yisong Fu ⋅ Zezhi Shao ⋅ Chengqing Yu ⋅ Yujie Li ⋅ Yongjun Xu ⋅ Xueqi Cheng ⋅ Fei Wang

关键词：时序分析，基础模型

30 TelecomTS: A Multi-Modal Observability Dataset for Time Series and Language Analysis

链接：https://icml.cc/virtual/2026/poster/61690

arXiv：http://arxiv.org/abs/2510.06063v1

代码：https://github.com/Ali-maatouk/TelecomTS

作者：Austin Feng ⋅ Andreas Varvarigos ⋅ Ioannis Panitsas ⋅ Daniela Fernandez ⋅ Yuwei Guo ⋅ Jinbiao Wei ⋅ Chen ⋅ Ali Maatouk ⋅ Leandros Tassiulas ⋅ ZHITAO YING

关键词：多模态可观测性数据集，通讯时序

31 Time Series, Vision, and Language: Exploring the Limits of Alignment in Contrastive Representation Spaces

链接：https://icml.cc/virtual/2026/poster/63821

arXiv：http://arxiv.org/abs/2602.19367v1

作者：Pratham Yashwante ⋅ Rose Yu

关键词：表示学习，三模态（时序，视觉，文本）对齐

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：2026-05-11，如有侵权请联系 cloudcommunity@tencent.com 删除

time

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