ICML 2026 | 时间序列（Time Series）论文总结（3）【因果，可解释性，不规则时序，表示学习等】

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3篇，共计39篇。

本文时间序列Topic：因果，可解释性，不规则时序，表示学习，benchmark等。

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 Credibility-Aware Weighting Federated Causal Discovery for Time Series

链接：https://icml.cc/virtual/2026/poster/62849

作者：Jiegang Xu ⋅ Fuyuan CAO ⋅ Jiye Liang

关键词：因果发现，可信度感知

2 Generalizing Multi-Scale Time-Series Modeling with a Single Operator

链接：https://icml.cc/virtual/2026/poster/62286

作者：Cheonwoo Lee ⋅ Dooho Lee ⋅ Doyun Choi ⋅ Jaemin Yoo

关键词：多尺度时序建模，核函数

3 BioFormer: Rethinking Cross-Subject Generalization via Spectral Structural Alignment in Biomedical Time-Series

链接：https://icml.cc/virtual/2026/poster/66155

作者：Du guikang ⋅ Haoran Li ⋅ Xinyu Liu ⋅ Zhibo Zhang ⋅ Xiaoli Gong ⋅ Jin Zhang

关键词：生物医学时序，谱漂移

4 Latent Laplace Diffusion for Irregular Multivariate Time Series

链接：https://icml.cc/virtual/2026/poster/61165

作者：Zinuo You ⋅ Jin Zheng ⋅ John Cartlidge

关键词：不规则时序，扩散模型，预测/插补

5 TiMi: Empower Time Series Transformers with Multimodal Mixture of Experts

链接：https://icml.cc/virtual/2026/poster/64391

arXiv：https://arxiv.org/abs/2602.21693

作者：Jiafeng Lin ⋅ Yuxuan Wang ⋅ HUAKUN LUO ⋅ Jianmin Wang ⋅ Zhongyi Pei

关键词：多模态时序，MoE

6 Towards a Unified Generative Model for Scarce Time Series with Domain Experts

链接：https://icml.cc/virtual/2026/poster/62223

作者：Zihao Yao ⋅ Qi Zheng ⋅ Jiankai Zuo ⋅ YAYING ZHANG

关键词：生成式，领域专家

7 Sonar-TS: Search-Then-Verify Natural Language Querying for Time Series Databases

链接：https://icml.cc/virtual/2026/poster/61266

arXiv：http://arxiv.org/abs/2602.17001v1

作者：Zhao Tan ⋅ Yiji Zhao ⋅ Shiyu Wang ⋅ Chang Xu ⋅ Yuxuan Liang ⋅ Xiping Liu ⋅ Shirui Pan ⋅ Ming Jin

关键词：时序数据库自然语言查询，神经符号框架

8 DualTimesField: Rethinking Time Series as Continuous-Time Trends and Events

链接：https://icml.cc/virtual/2026/poster/60845

作者：Wencheng Zhang ⋅ Long Li ⋅ Huayi Qin ⋅ Zongjuan Wu ⋅ Jing Li ⋅ Wanghu Chen

关键词：时序表示学习，神经场

9 A Spiking Heterogeneous Harmonic Resonate-and-Fire State Space Model for Time Series

链接：https://icml.cc/virtual/2026/poster/62795

作者：Kartikay Agrawal ⋅ Vaishnavi Nagabhushana ⋅ Abhijeet Vikram ⋅ Vedant Sharma ⋅ Ayon Borthakur

关键词：脉冲神经网络，状态空间

10 TopoDistill: Distilling Global System Topology for Causal Discovery in Multivariate Time Series

链接：https://icml.cc/virtual/2026/poster/64953

作者：Zehao Liu ⋅ Pengfei Jiao ⋅ Yuhan Wu ⋅ Jianqi Yang ⋅ Yuyu Yin

关键词：因果推断，知识蒸馏

11 Learning Fingerprints for Medical Time Series with Redundancy-Constrained Information Maximization

链接：https://icml.cc/virtual/2026/poster/61898

arXiv：http://arxiv.org/abs/2605.00130v1

作者：Huayu Li ⋅ ZhengXiao He ⋅ Xiwen Chen ⋅ Jingjing Wang ⋅ Siyuan Tian ⋅ Jinghao Wen ⋅ Ao Li

关键词：解耦率失真，医疗时序

12 Dynamic Relational Priming Improves Transformer in Multivariate Time Series

链接：https://icml.cc/virtual/2026/poster/62910

arXiv：http://arxiv.org/abs/2509.12196v1

作者：Hunjae Lee ⋅ Corey Clark

关键词：注意力机制，通道关系建模

13 Function-Valued Causal Influence in Nonlinear Time Series

链接：https://icml.cc/virtual/2026/poster/64652

作者：Valentina Kuskova ⋅ Dmitry Zaytsev ⋅ Michael Coppedge

关键词：因果发现，函数值因果影响

14 Position: Why a Dynamical Systems Perspective is Needed to Advance Time Series Modeling

链接：https://icml.cc/virtual/2026/poster/67104

arXiv：http://arxiv.org/abs/2602.16864v1

作者：Daniel Durstewitz ⋅ Christoph Jürgen Hemmer ⋅ Florian Hess ⋅ Charlotte Ricarda Doll ⋅ Lukas Eisenmann

关键词：动力系统

15 PATRA: Pattern-Aware Alignment and Balanced Reasoning for Time Series Question Answering

链接：https://icml.cc/virtual/2026/poster/65414

arXiv：http://arxiv.org/abs/2602.23161v1

作者：Junkai Lu ⋅ Peng Chen ⋅ Xingjian Wu ⋅ Yang Shu ⋅ Chenjuan Guo ⋅ Christian S Jensen ⋅ Bin Yang

关键词：问答，对齐，推理

16 Giving Sensors a Voice: Multimodal JEPA for Semantic Time-Series Embeddings

链接：https://icml.cc/virtual/2026/poster/61949

作者：Gerardo Pastrana ⋅ Sina Pakazad ⋅ Henrik Ohlsson ⋅ Utsav Dutta

关键词：通道感知，时序嵌入

17 Causal Discovery for Irregularly Time Series with Consistency Guarantees

链接：https://icml.cc/virtual/2026/poster/60649

作者：Weihong Li ⋅ Baohong Li ⋅ Anpeng Wu ⋅ Zhihan Li ⋅ Ming Ma ⋅ Kun Kuang ⋅ Keting Yin

关键词：因果发现

18 Robust Causal Discovery in Real-World Time Series with Power-Laws

链接：https://icml.cc/virtual/2026/poster/66032

arXiv：http://arxiv.org/abs/2507.12257v3

作者：Matteo Tusoni ⋅ Giuseppe Masi ⋅ Andrea Coletta ⋅ Aldo Glielmo ⋅ Viviana Arrigoni ⋅ Novella Bartolini

关键词：因果发现，稳健性

19 Causal discovery for time series with endogenous context variables

链接：https://icml.cc/virtual/2026/poster/66592

作者：Oana-Iuliana Popescu ⋅ Wiebke Günther ⋅ Martin Rabel ⋅ Jakob Runge

关键词：因果发现，结构因果模型

20 Time series saliency maps: Explaining models across multiple domains

链接：https://icml.cc/virtual/2026/poster/65637

arXiv：http://arxiv.org/abs/2505.13100v3

作者：Christodoulos Kechris ⋅ Jonathan Dan ⋅ David Atienza

关键词：模型解释

21 TimeSAE: Sparse Decoding for Faithful Explanations of Black-Box Time Series Models

链接：https://icml.cc/virtual/2026/poster/66052

arXiv：http://arxiv.org/abs/2601.09776v1

作者：Khalid Oublal ⋅ Quentin Bouniot ⋅ Qi Gan ⋅ Stephan Clemencon ⋅ Zeynep Akata

关键词：稀疏解码，公平解释

22 TSRBench: A Comprehensive Multi-task Multi-modal Time Series Reasoning Benchmark for Generalist Models

链接：https://icml.cc/virtual/2026/poster/63969

arXiv：http://arxiv.org/abs/2601.18744v1

代码：https://github.com/tianyi-lab/TSRBench

作者：Fangxu Yu ⋅ Xingang Guo ⋅ Lingzhi Yuan ⋅ Haoqiang Kang ⋅ Hongyu Zhao ⋅ Lianhui Qin ⋅ Furong Huang ⋅ Bin Hu ⋅ Tianyi Zhou

关键词：多模态时序推理

23 ProSAR: Prototype-Guided Semantic Augmentation and Refinement for Time Series Contrastive Learning

链接：https://icml.cc/virtual/2026/poster/62342

作者：Caiyi Yang ⋅ Chenglin Li ⋅ Hao Zhang ⋅ Weijia Lu ⋅ ZHIFEI YANG ⋅ Wenrui Dai ⋅ xiaodong Zhang ⋅ Xiaofeng Ma ⋅ Can Zhang ⋅ Junni Zou ⋅ Hongkai Xiong

关键词：对比学习，语义增强

24 Terminal Dimension Reduction for Time Series with Applications

链接：https://icml.cc/virtual/2026/poster/62402

作者：Alexander Munteanu ⋅ Matteo Russo ⋅ David Saulpic ⋅ Chris Schwiegelshohn

关键词：终端嵌入

25 NeurOCNN: A Neural-Operator-Based Model for Physiological Time Series

链接：https://icml.cc/virtual/2026/poster/62493

代码：https://github.com/dcoder444/NeurOCNN

作者：Daya Kumar ⋅ Uday Devulapalli ⋅ Aarat Satsangi ⋅ Apurva Narayan

关键词：神经算子，生理时序

26 TimeLAVA: Learning-Agnostic Valuation for Time Series Data

链接：https://icml.cc/virtual/2026/poster/62497

作者：Wenqin Liu ⋅ Weizhi Quan ⋅ Aoqi Zuo ⋅ Erdun Gao ⋅ Vu Nguyen ⋅ Dino Sejdinovic ⋅ Howard Bondell ⋅ Mingming Gong

关键词：数据估值，异常检测，数据剪枝，标签噪声检测

27 Ranking Time Series using a Time Warping Ideal Point Model

链接：https://icml.cc/virtual/2026/poster/66190

作者：Lucas Zoroddu ⋅ Pierre Humbert ⋅ Laurent Oudre

关键词：排序，DTW

28 ZeroDiff: Zero-Shot Time Series Reconstruction via Informed-Prior Diffusion

链接：https://icml.cc/virtual/2026/poster/66272

作者：Yingda Fan ⋅ Dan Lu ⋅ Xiaowei Jia

关键词：零样本时序重构

29 Generative Modeling of Irregular Time Series via SDE-Induced Continuous-Discrete Variational Inference

链接：https://icml.cc/virtual/2026/poster/64934

作者：Zexin Yuan ⋅ Qinliang Su ⋅ Junxi Xiao

关键词：不规则时序，随机微分方程，变分推断

30 Position: Interpretability in Deep Time Series Models Demands Semantic Alignment

链接：https://icml.cc/virtual/2026/poster/67131

作者：Giovanni De Felice ⋅ Riccardo D`Elia ⋅ Alberto Termine ⋅ Pietro Barbiero ⋅ Giuseppe Marra ⋅ Silvia Santini

关键词：语义对齐，可解释性

31 QuITE: Query-based Irregular Time-series Embedding

链接：https://icml.cc/virtual/2026/poster/64962

作者：Junghoon Lim

关键词：不规则时序，查询

32 Exposing Vulnerabilities in Explanation for Time Series Classifiers via Dual-Target Attacks

链接：https://icml.cc/virtual/2026/poster/66770

arXiv：http://arxiv.org/abs/2602.02763v2

作者：Bohan Wang ⋅ Zewen Liu ⋅ Lu Lin ⋅ Hui Liu ⋅ Li Xiong ⋅ Ming Jin ⋅ Wei Jin

关键词：可解释性检测

33 Unified Time Series Explanations via Semi-Amortized Optimization and Instance-level Multi-Expert Knowledge Distillation

链接：https://icml.cc/virtual/2026/poster/62730

作者：Viet-Hung Tran ⋅ Zichi Zhang ⋅ Ngoc Doan ⋅ Xuan Hoang Nguyen ⋅ Phi Nguyen ⋅ Yimeng An ⋅ Peixin Li ⋅ Hans Vandierendonck ⋅ Ira Assent ⋅ Son Thai

关键词：可解释性，知识蒸馏

34 Adaptive Time Series Reasoning via Segment Selection

链接：https://icml.cc/virtual/2026/poster/60563

arXiv：http://arxiv.org/abs/2602.18645v1

作者：Shvat Messica ⋅ Jiawen Zhang ⋅ Kevin Li ⋅ Theodoros Tsiligkaridis ⋅ Marinka Zitnik

关键词：推理，片段选择

35 Words Towards Explainability: Caption Label-Free Learning via Dual Loop Agentic Time SeriesCaptioning

链接：https://icml.cc/virtual/2026/poster/64438

作者：Difei Hou ⋅ Jiaqi Yue ⋅ Chunhui Zhao

关键词：时间序列文本描述生成，可解释性

36 ReAugment: Targeted Few-Shot Time Series Augmentation via Model Zoo-Guided Reinforcement Learning

链接：https://icml.cc/virtual/2026/poster/63111

作者：Haochen Yuan ⋅ Yutong Wang ⋅ Yihong Chen ⋅ Yunbo Wang ⋅ Xiaokang Yang

关键词：小样本时序预测，强化学习，数据增广

37 Sparse Regression with Constraints for -Mixing Time Series: Algorithms and Guarantees

链接：https://icml.cc/virtual/2026/poster/63855

作者：Ruoxin Yuan ⋅ Lijun Ding

关键词：α- 混合平稳高斯过程，高斯向量自回归

38 BEDTime: A Unified Benchmark for Automatically Describing Time Series

链接：https://icml.cc/virtual/2026/poster/63965

作者：Medhasweta Sen ⋅ Zachary Gottesman ⋅ Jiaxing Qiu ⋅ C. Bayan Bruss ⋅ Nam Nguyen ⋅ Thomas Hartvigsen

关键词：benchmark，时序结构特征描述

39 Time-Series Decomposition as a standalone Task: A Mechanism-Driven Diagnostic Benchmark

链接：https://icml.cc/virtual/2026/poster/63747

作者：Zipeng Wu ⋅ Jiani Wei ⋅ Shiqiao Zhou ⋅ Jiajun Chen ⋅ Fabian Spill ⋅ J. Andrews

关键词：时序分解，benchmark

如果觉得有帮助还请分享，在看，点赞