Zezhi Shao 邵泽志

* 通讯作者；† 共同第一作者。

1. KDD

   Pre-training Enhanced Spatial-temporal Graph Neural Network for Multivariate Time Series Forecasting

   Zezhi Shao, Zhao Zhang, Fei Wang, and Yongjun Xu

   In Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, Washington DC, USA, 2022

   The 3rd-Most Cited paper in KDD 2022 摘要 DOI 引用 PDF 代码

   

   The 3rd-Most Cited paper in KDD 2022 3/254

   Multivariate Time Series (MTS) forecasting plays a vital role in a wide range of applications. Recently, Spatial-Temporal Graph Neural Networks (STGNNs) have become increasingly popular MTS forecasting methods. STGNNs jointly model the spatial and temporal patterns of MTS through graph neural networks and sequential models, significantly improving the prediction accuracy. But limited by model complexity, most STGNNs only consider short-term historical MTS data, such as data over the past one hour. However, the patterns of time series and the dependencies between them (i.e., the temporal and spatial patterns) need to be analyzed based on long-term historical MTS data. To address this issue, we propose a novel framework, in which STGNN is Enhanced by a scalable time series Pre-training model (STEP). Specifically, we design a pre-training model to efficiently learn temporal patterns from very long-term history time series (e.g., the past two weeks) and generate segment-level representations. These representations provide contextual information for short-term time series input to STGNNs and facilitate modeling dependencies between time series. Experiments on three public real-world datasets demonstrate that our framework is capable of significantly enhancing downstream STGNNs, and our pre-training model aptly captures temporal patterns.

   @inproceedings{10.1145/3534678.3539396,
     author = {Shao, Zezhi and Zhang, Zhao and Wang, Fei and Xu, Yongjun},
     title = {Pre-training Enhanced Spatial-temporal Graph Neural Network for Multivariate Time Series Forecasting},
     year = {2022},
     month = aug,
     isbn = {9781450393850},
     publisher = {Association for Computing Machinery},
     address = {New York, NY, USA},
     url = {https://doi.org/10.1145/3534678.3539396},
     doi = {10.1145/3534678.3539396},
     booktitle = {Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining},
     pages = {1567-1577},
     numpages = {11},
     keywords = {multivariate time series forecasting, pre-training model, spatial-temporal graph neural network},
     location = {Washington DC, USA},
     series = {KDD '22},
   }

2. VLDB

   Decoupled Dynamic Spatial-Temporal Graph Neural Network for Traffic Forecasting

   Zezhi Shao, Zhao Zhang, Wei Wei, Fei Wang, Yongjun Xu, Xin Cao, and Christian S. Jensen

   Proc. VLDB Endow., 2022

   ESI Highly Cited Papers | The 3rd-Most Cited paper in VLDB 2022 摘要 DOI 引用 PDF 代码

   

   The 3rd-Most Cited paper in VLDB 2022 3/357

   We all depend on mobility, and vehicular transportation affects the daily lives of most of us. Thus, the ability to forecast the state of traffic in a road network is an important functionality and a challenging task. Traffic data is often obtained from sensors deployed in a road network. Recent proposals on spatial-temporal graph neural networks have achieved great progress at modeling complex spatial-temporal correlations in traffic data, by modeling traffic data as a diffusion process. However, intuitively, traffic data encompasses two different kinds of hidden time series signals, namely the diffusion signals and inherent signals. Unfortunately, nearly all previous works coarsely consider traffic signals entirely as the outcome of the diffusion, while neglecting the inherent signals, which impacts model performance negatively. To improve modeling performance, we propose a novel Decoupled Spatial-Temporal Framework (DSTF) that separates the diffusion and inherent traffic information in a data-driven manner, which encompasses a unique estimation gate and a residual decomposition mechanism. The separated signals can be handled subsequently by the diffusion and inherent modules separately. Further, we propose an instantiation of DSTF, Decoupled Dynamic Spatial-Temporal Graph Neural Network (D2STGNN), that captures spatial-temporal correlations and also features a dynamic graph learning module that targets the learning of the dynamic characteristics of traffic networks. Extensive experiments with four real-world traffic datasets demonstrate that the framework is capable of advancing the state-of-the-art.

   @article{10.14778/3551793.3551827,
     author = {Shao, Zezhi and Zhang, Zhao and Wei, Wei and Wang, Fei and Xu, Yongjun and Cao, Xin and Jensen, Christian S.},
     title = {Decoupled Dynamic Spatial-Temporal Graph Neural Network for Traffic Forecasting},
     year = {2022},
     publisher = {VLDB Endowment},
     volume = {15},
     number = {11},
     issn = {2150-8097},
     url = {https://doi.org/10.14778/3551793.3551827},
     doi = {10.14778/3551793.3551827},
     journal = {Proc. VLDB Endow.},
     month = jul,
     pages = {2733-2746},
     numpages = {14},
   }

3. TKDE

   Exploring Progress in Multivariate Time Series Forecasting: Comprehensive Benchmarking and Heterogeneity Analysis

   Zezhi Shao, Fei Wang, Yongjun Xu^*, Wei Wei, Chengqing Yu, Zhao Zhang, Di Yao, Tao Sun, Guangyin Jin, Xin Cao ， 另 3 位作者

   IEEE Transactions on Knowledge and Data Engineering, 2025

   ESI Hot Papers | ESI Highly Cited Papers | Acquired 1.8k+ Stars at Github 摘要 DOI 引用 PDF 代码

   

   BasicTS+ has entered ESI Highly Cited Papers and acquired 1.8+ Stars for the First, one of the Most Popular, fair and scalable benchmark of timeseries forecasting*

   Multivariate Time Series (MTS) analysis is crucial to understanding and managing complex systems, such as traffic and energy systems, and a variety of approaches to MTS forecasting have been proposed recently. However, we often observe inconsistent or seemingly contradictory performance findings across different studies. This hinders our understanding of the merits of different approaches and slows down progress. We address the need for means of assessing MTS forecasting proposals reliably and fairly, in turn enabling better exploitation of MTS as seen in different applications. Specifically, we first propose BasicTS+, a benchmark designed to enable fair, comprehensive, and reproducible comparison of MTS forecasting solutions. BasicTS+ establishes a unified training pipeline and reasonable settings, enabling an unbiased evaluation. Second, we identify the heterogeneity across different MTS as an important consideration and enable classification of MTS based on their temporal and spatial characteristics. Disregarding this heterogeneity is a prime reason for difficulties in selecting the most promising technical directions. Third, we apply BasicTS+ along with rich datasets to assess the capabilities of more than 30 MTS forecasting solutions. This provides readers with an overall picture of the cutting-edge research on MTS forecasting.

   @article{10726722,
     author = {Shao, Zezhi and Wang, Fei and Xu, Yongjun and Wei, Wei and Yu, Chengqing and Zhang, Zhao and Yao, Di and Sun, Tao and Jin, Guangyin and Cao, Xin and Cong, Gao and Jensen, Christian S. and Cheng, Xueqi},
     journal = {IEEE Transactions on Knowledge and Data Engineering},
     title = {Exploring Progress in Multivariate Time Series Forecasting: Comprehensive Benchmarking and Heterogeneity Analysis},
     year = {2025},
     volume = {37},
     number = {1},
     pages = {291-305},
     keywords = {Forecasting;Time series analysis;Benchmark testing;Transformers;Predictive models;Data models;Computer science;Reliability;Proposals;Electricity;Benchmarking;multivariate time series;spatial-temporal forecasting;long-term time series forecasting},
     doi = {10.1109/TKDE.2024.3484454},
   }

4. CIKM

   Spatial-Temporal Identity: A Simple yet Effective Baseline for Multivariate Time Series Forecasting

   Zezhi Shao, Zhao Zhang, Fei Wang, Wei Wei, and Yongjun Xu

   In Proceedings of the 31st ACM International Conference on Information & Knowledge Management, Atlanta, GA, USA, 2022

   The Most Cited paper in CIKM 2022 摘要 DOI 引用 PDF 代码

   

   The Most Cited paper in CIKM 2022 1/561

   Multivariate Time Series (MTS) forecasting plays a vital role in a wide range of applications. Recently, Spatial-Temporal Graph Neural Networks (STGNNs) have become increasingly popular MTS forecasting methods due to their state-of-the-art performance. However, recent works are becoming more sophisticated with limited performance improvements. This phenomenon motivates us to explore the critical factors of MTS forecasting and design a model that is as powerful as STGNNs, but more concise and efficient. In this paper, we identify the indistinguishability of samples in both spatial and temporal dimensions as a key bottleneck, and propose a simple yet effective baseline for MTS forecasting by attaching Spatial and Temporal IDentity information (STID), which achieves the best performance and efficiency simultaneously based on simple Multi-Layer Perceptrons (MLPs). These results suggest that we can design efficient and effective models as long as they solve the indistinguishability of samples, without being limited to STGNNs.

   @inproceedings{10.1145/3511808.3557702,
     author = {Shao, Zezhi and Zhang, Zhao and Wang, Fei and Wei, Wei and Xu, Yongjun},
     title = {Spatial-Temporal Identity: A Simple yet Effective Baseline for Multivariate Time Series Forecasting},
     year = {2022},
     month = oct,
     isbn = {9781450392365},
     publisher = {Association for Computing Machinery},
     address = {New York, NY, USA},
     url = {https://doi.org/10.1145/3511808.3557702},
     doi = {10.1145/3511808.3557702},
     booktitle = {Proceedings of the 31st ACM International Conference on Information \& Knowledge Management},
     pages = {4454-4458},
     numpages = {5},
     keywords = {spatial-temporal graph neural network, multivariate time series forecasting, baseline},
     location = {Atlanta, GA, USA},
     series = {CIKM '22},
   }

5. VLDB

   QUALS: Corpus Equilibrium for Universal Forecasting via Pattern Quantization and Learnability Synchronization

   Yujie Li, Zezhi Shao^*, Chengqing Yu, Yisong Fu, Weijie Zhu, Yifan Du, Jilin Hu, Bin Yang, Yongjun Xu, and Fei Wang

   Proc. VLDB Endow., 2027

   Accepted in September 2026

   DOI arXiv 引用 PDF 代码

   

   @article{li2027quals,
     title = {QUALS: Corpus Equilibrium for Universal Forecasting via Pattern Quantization and Learnability Synchronization},
     author = {Li, Yujie and Shao, Zezhi and Yu, Chengqing and Fu, Yisong and Zhu, Weijie and Du, Yifan and Hu, Jilin and Yang, Bin and Xu, Yongjun and Wang, Fei},
     journal = {Proc. VLDB Endow.},
     year = {2027},
     note = {Accepted in September 2026},
     doi = {10.48550/arXiv.2609.20156},
     url = {https://arxiv.org/abs/2609.20156},
   }

6. TKDE

   Spatio-Temporal Graph Neural Networks for Predictive Learning in Urban Computing: A Survey

   Guangyin Jin, Yuxuan Liang, Yuchen Fang, Zezhi Shao, Jincai Huang, Junbo Zhang, and Yu Zheng

   IEEE Transactions on Knowledge and Data Engineering, 2024

   ESI Hot Papers | ESI Highly Cited Papers DOI 引用 PDF

   

   This paper has entered ESI Hot Papers.

   @article{DBLP:journals/tkde/JinLFSHZZ24,
     author = {Jin, Guangyin and Liang, Yuxuan and Fang, Yuchen and Shao, Zezhi and Huang, Jincai and Zhang, Junbo and Zheng, Yu},
     journal = {IEEE Transactions on Knowledge and Data Engineering},
     title = {Spatio-Temporal Graph Neural Networks for Predictive Learning in Urban Computing: A Survey},
     year = {2024},
     month = nov,
     volume = {36},
     number = {10},
     pages = {5388-5408},
     keywords = {spatio-temporal data mining, graph neural networks, urban computing, predictive learning, time series},
     doi = {10.1109/TKDE.2023.3333824},
   }

7. KDD

   BLAST: Balanced Sampling Time Series Corpus for Universal Forecasting Models

   Zezhi Shao, Yujie Li, Fei Wang, Chengqing Yu, Yisong Fu, Tangwen Qian, Bin Xu, Boyu Diao, Yongjun Xu, and Xueqi Cheng

   In Proceedings of the 31th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, Toronton, ON, Canada, 2025

   摘要 引用 PDF 代码

   

   The advent of universal time series forecasting models has revolutionized zero-shot forecasting across diverse domains, yet the critical role of data diversity in training these models remains underexplored. Existing large-scale time series datasets often suffer from inherent biases and imbalanced distributions, leading to suboptimal model performance and generalization. To address this gap, we introduce BLAST, a novel pre-training corpus designed to enhance data diversity through a balanced sampling strategy. First, BLAST incorporates 321 billion observations from publicly available datasets and employs a comprehensive suite of statistical metrics to characterize time series patterns. Then, to facilitate pattern-oriented sampling, the data is implicitly clustered using grid-based partitioning. Furthermore, by integrating grid sampling and grid mixup techniques, BLAST ensures a balanced and representative coverage of diverse patterns. Experimental results demonstrate that models pre-trained on BLAST achieve state-of-the-art performance with a fraction of the computational resources and training tokens required by existing methods. Our findings highlight the pivotal role of data diversity in improving both training efficiency and model performance for the universal forecasting task.

   @inproceedings{blast,
     author = {Shao, Zezhi and Li, Yujie and Wang, Fei and Yu, Chengqing and Fu, Yisong and Qian, Tangwen and Xu, Bin and Diao, Boyu and Xu, Yongjun and Cheng, Xueqi},
     title = {BLAST: Balanced Sampling Time Series Corpus for Universal Forecasting Models},
     year = {2025},
     publisher = {Association for Computing Machinery},
     address = {New York, NY, USA},
     booktitle = {Proceedings of the 31th ACM SIGKDD Conference on Knowledge Discovery and Data Mining},
     keywords = {large-scale time series dataset, balanced sampling, universal time series forecasting},
     location = {Toronton, ON, Canada},
     series = {KDD '25},
   }

8. TKDE

   Heterogeneous Graph Neural Network With Multi-View Representation Learning

   Zezhi Shao, Yongjun Xu, Wei Wei, Fei Wang, Zhao Zhang, and Feida Zhu

   IEEE Transactions on Knowledge and Data Engineering, 2023

   DOI 引用 PDF

   

   @article{9961953,
     author = {Shao, Zezhi and Xu, Yongjun and Wei, Wei and Wang, Fei and Zhang, Zhao and Zhu, Feida},
     journal = {IEEE Transactions on Knowledge and Data Engineering},
     title = {Heterogeneous Graph Neural Network With Multi-View Representation Learning},
     year = {2023},
     month = nov,
     volume = {35},
     number = {11},
     pages = {11476-11488},
     keywords = {Semantics;Mercury (metals);Graph neural networks;Aggregates;Task analysis;Representation learning;Adaptation models;Heterogeneous graphs;graph neural networks;graph embedding},
     doi = {10.1109/TKDE.2022.3224193},
   }

9. Innovation

   AI for science: Progress, challenges, and perspectives

   Yongjun Xu, Zezhi Shao^†, Xin Liu, Yuquan Li, Zhulin An, Chenguang Fu, Changping Huang, Lin Zhang, Feng Gao, Jian Li ， 另 130 位作者

   The Innovation, 2026

   Accepted in August 2026

   DOI 引用 HTML

   

   @article{xu2026aiforscience,
     title = {AI for science: Progress, challenges, and perspectives},
     author = {Xu, Yongjun and Shao, Zezhi and Liu, Xin and Li, Yuquan and An, Zhulin and Fu, Chenguang and Huang, Changping and Zhang, Lin and Gao, Feng and Li, Jian and Li, Yuyang and Yao, Fanglong and Xia, Anquan and Chen, Kaihua and Zheng, Xiaolong and Yang, Chuanguang and Huang, Libo and Dong, Fengliang and Yao, Shanliang and Zhang, Zhao and Wang, Jue and Wang, Qi and Wang, Yuanzhuo and Li, Haisheng and Xiao, Nachuan and Wang, Yuyang and Feng, Guoshuang and Lai, Puxiang and Song, Yuxuan and Xu, Tao and Yang, Jie and Peng, Yun and Meng, Xiangbin and Peng, Xin and Li, Kunjie and Niu, Yupeng and Xu, Tiancheng and Xu, Huiyu and Huang, Tao and Han, Yong and Li, Min and Hou, Jiayun and Lu, Xiaoling and Yan, Fangrong and Zhao, Yang and Wei, Yongyue and Yin, Ming-Jie and Wang, Zongguo and Xiang, Leilei and Qiu, Yubao and Zeng, Jiangyuan and Liu, Yang and Chang, Yue and Grunwald, Sabine and Peijnenburg, Willie and Hofmann, Thilo and Roepman, Ronald and Li, Jun and Dercon, Gerd and Gao, Lianru and Vezzone, Mariana and Virta, Marko and Dietmann, Sabine and Yuan, Deshuai and Chen, Fei and Chen, Wei and Wu, Kai and Chen, Tao and Zhang, Haibo and Yuan, Zuoqiang and Han, Jifeng and Zhu, Xuetao and Zhang, Bin and Su, Wentao and Meng, Wei and Zhang, Ming and Zhang, Dachuan and Wu, Qingyan and Lin, Xiaojie and Peng, Xiayao and Yang, Tao and Wu, Wei and Gan, Lei and Wan, Meng and Ding, Ye and Wei, Zhongbao and Yang, Shaohua and Zhang, Yunfei and Yang, Lijun and Yu, Kun and Duan, Yuanyuan and Chen, Xingying and Jia, Peng and Luo, Ali and Lin, Jiaben and Wang, Cunshi and Zhang, Zhiting and Liu, Xinhe and Zhu, Dongchen and Hu, Peng and Li, Tong and Wu, Jinran and Dalal, Ram and Wu, Yang and Yu, Chengqing and Li, Zixuan and Zhu, Weijie and Lu, Tianyou and Yun, Lijun and Jelinek, Thorsten and Wang, Wenguan and Liu, Tiange and Wang, Fei and Li, Heyang and Di, Shimin and Diao, Boyu and Wang, Xiangfeng and Cao, Xin and Yin, Zhigang and Wang, Fang and Zhang, Dake and Luo, Ming and Qian, Sen and Yu, Ping and Liu, Xingchen and Shen, Jun and Li, Zhengyang and Dai, Ji and Lin, Yi and Wang, You-Gan and McLachlan, Geoffrey John and Missous, Mohamed and Rillig, Matthias C. and Naidu, Ravi and Amelung, Wulf and Tiedje, James M. and Shi, Chuang and Wang, Yanfen and Zhang, Bing and Zhang, Jiabao},
     journal = {The Innovation},
     year = {2026},
     month = aug,
     pages = {101530},
     note = {Accepted in August 2026},
     doi = {10.1016/j.xinn.2026.101530},
     url = {https://doi.org/10.1016/j.xinn.2026.101530},
   }

10. KDD

    Efficient large-scale traffic forecasting with transformers: A spatial data management perspective

    Yuchen Fang, Yuxuan Liang, Bo Hui, Zezhi Shao, Liwei Deng, Xu Liu, Xinke Jiang, and Kai Zheng

    In Proceedings of the 31th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, Toronton, ON, Canada, 2025

    Most Influential KDD 2025 Papers (PaperDigest) 摘要 引用 PDF 代码

    

    Most Influential KDD 2025 Papers (PaperDigest)

    Road traffic forecasting is crucial in real-world intelligent transportation scenarios like traffic dispatching and path planning in city management and personal traveling. Spatio-temporal graph neural networks (STGNNs) stand out as the mainstream solution in this task. Nevertheless, the quadratic complexity of remarkable dynamic spatial modeling-based STGNNs has become the bottleneck over large-scale traffic data. From the spatial data management perspective, we present a novel Transformer framework called PatchSTG to efficiently and dynamically model spatial dependencies for large-scale traffic forecasting with interpretability and fidelity. Specifically, we design a novel irregular spatial patching to reduce the number of points involved in the dynamic calculation of Transformer. The irregular spatial patching first utilizes the leaf K-dimensional tree (KDTree) to recursively partition irregularly distributed traffic points into leaf nodes with a small capacity, and then merges leaf nodes belonging to the same subtree into occupancy-equaled and non-overlapped patches through padding and backtracking. Based on the patched data, depth and breadth attention are used interchangeably in the encoder to dynamically learn local and global spatial knowledge from points in a patch and points with the same index of patches. Experimental results on four real world large-scale traffic datasets show that our PatchSTG achieves train speed and memory utilization improvements up to 10x and 4x with the state-of-the-art performance.

    @inproceedings{patchstg,
      author = {Fang, Yuchen and Liang, Yuxuan and Hui, Bo and Shao, Zezhi and Deng, Liwei and Liu, Xu and Jiang, Xinke and Zheng, Kai},
      title = {Efficient large-scale traffic forecasting with transformers: A spatial data management perspective},
      year = {2025},
      publisher = {Association for Computing Machinery},
      address = {New York, NY, USA},
      booktitle = {Proceedings of the 31th ACM SIGKDD Conference on Knowledge Discovery and Data Mining},
      keywords = {traffic forecasting, Transformer, spatial data management},
      location = {Toronton, ON, Canada},
      series = {KDD '25},
    }

11. Innovation

    Spatial-temporal large models: A super hub linking multiple scientific areas with artificial intelligence

    Zezhi Shao, Tangwen Qian, Tao Sun, Fei Wang, and Yongjun Xu^*

    The Innovation, 2025

    DOI 引用 PDF

    

    @article{SHAO2025100763,
      title = {Spatial-temporal large models: A super hub linking multiple scientific areas with artificial intelligence},
      journal = {The Innovation},
      volume = {6},
      number = {2},
      pages = {100763},
      year = {2025},
      issn = {2666-6758},
      doi = {https://doi.org/10.1016/j.xinn.2024.100763},
      url = {https://www.sciencedirect.com/science/article/pii/S2666675824002017},
      author = {Shao, Zezhi and Qian, Tangwen and Sun, Tao and Wang, Fei and Xu, Yongjun},
    }

12. COMMTR

    HUTFormer: Hierarchical U-Net transformer for long-term traffic forecasting

    Zezhi Shao, Fei Wang, Tao Sun, Chengqing Yu, Yuchen Fang, Guangyin Jin, Zhulin An, Yang Liu, Xiaobo Qu, and Yongjun Xu

    Communications in Transportation Research, 2025

    摘要 DOI PDF

    

    Traffic forecasting, which aims to predict traffic conditions based on historical observations, has been an enduring research topic and is widely recognized as an essential component of intelligent transportation. Recent proposals on Spatial-Temporal Graph Neural Networks (STGNNs) have made significant progress by combining sequential models with graph convolution networks. However, due to high complexity issues, STGNNs only focus on short-term traffic forecasting (e.g., 1-h ahead), while ignoring more practical long-term forecasting. In this paper, we make the first attempt to explore long-term traffic forecasting (e.g., 1-day ahead). To this end, we first reveal its unique challenges in exploiting multi-scale representations. Then, we propose a novel Hierarchical U-net TransFormer (HUTFormer) to address the issues of long-term traffic forecasting. HUTFormer consists of a hierarchical encoder and decoder to jointly generate and utilize multi-scale representations of traffic data. Specifically, for the encoder, we propose window self-attention and segment merging to extract multi-scale representations from long-term traffic data. For the decoder, we design a cross-scale attention mechanism to effectively incorporate multi-scale representations. In addition, HUTFormer employs an efficient input embedding strategy to address the complexity issues. Extensive experiments on four traffic datasets show that the proposed HUTFormer significantly outperforms state-of-the-art traffic forecasting and long time series forecasting baselines.

13. NeurIPS

    Selective Learning for Deep Time Series Forecasting

    Yisong Fu, Zezhi Shao, Chengqing Yu, Yujie Li, Zhulin An, Cheems Wang, Yongjun Xu, and Fei Wang

    In The Thirty-ninth Annual Conference on Neural Information Processing Systems, 2025

    摘要 DOI 引用 PDF 代码

    

    Benefiting from high capacity for capturing complex temporal patterns, deep learning (DL) has significantly advanced time series forecasting (TSF). However, deep models tend to suffer from severe overfitting due to the inherent vulnerability of time series to noise and anomalies. The prevailing DL paradigm uniformly optimizes all timesteps through the MSE loss and learns those uncertain and anomalous timesteps without difference, ultimately resulting in overfitting. To address this, we propose a novel selective learning strategy for deep TSF. Specifically, selective learning screens a subset of the whole timesteps to calculate the MSE loss in optimization, guiding the model to focus on generalizable timesteps while disregarding non-generalizable ones. Our framework introduces a dual-mask mechanism to target timesteps: (1) an uncertainty mask leveraging residual entropy to filter uncertain timesteps, and (2) an anomaly mask employing residual lower bound estimation to exclude anomalous timesteps. Extensive experiments across eight real-world datasets demonstrate that selective learning can significantly improve the predictive performance for typical state-of-the-art deep models, including 37.4% MSE reduction for Informer, 8.4% for TimesNet, and 6.5% for iTransformer.

    @inproceedings{fu2025selective,
      title = {Selective Learning for Deep Time Series Forecasting},
      booktitle = {The Thirty-ninth Annual Conference on Neural Information Processing Systems},
      year = {2025},
      url = {https://openreview.net/forum?id=kgzRy6nD6D},
      author = {Fu, Yisong and Shao, Zezhi and Yu, Chengqing and Li, Yujie and An, Zhulin and Wang, Cheems and Xu, Yongjun and Wang, Fei},
      doi = {https://doi.org/10.48550/arXiv.2510.25207},
    }

14. ICML

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

    Yisong Fu, Zezhi Shao, Chengqing Yu, Yujie Li, Yongjun Xu, Xueqi Cheng, and Fei Wang

    In The Forty-Third International Conference on Machine Learning, 2026

    摘要 DOI arXiv 引用 PDF 代码

    

    We present Zeus, a unified tuning-free Time Series Foundation Model (TSFM) that delivers superior performance across diverse analysis tasks without any task-specific fine-tuning. Unlike prior studies that primarily focus on zero-shot forecasting but require task-specific tuning for other tasks, Zeus bridges this gap by addressing two fundamental challenges in multi-task generalization. First, to reconcile point-level granularity with long-sequence scalability, Zeus incorporates a multi-scale Transformer featuring point-wise tokenization and a U-shaped hierarchy, effectively balancing fine-grained fidelity with computational efficiency. Second, to accommodate varying inductive biases across different tasks, Zeus introduces Multi-Objective Temporal Masking (MOTM), a unified strategy that supports heterogeneous tasks (e.g., extrapolation, interpolation, and global abstraction) within a single framework. Extensive experiments across five representative tasks demonstrate that Zeus consistently achieves competitive results in tuning-free settings, underscoring its potential as a general-purpose TSFM.

    @inproceedings{zeus,
      title = {Zeus: Towards Tuning-Free Foundation Model for Time Series Analysis},
      booktitle = {The Forty-Third International Conference on Machine Learning},
      year = {2026},
      url = {https://arxiv.org/abs/2607.01918},
      author = {Fu, Yisong and Shao, Zezhi and Yu, Chengqing and Li, Yujie and Xu, Yongjun and Cheng, Xueqi and Wang, Fei},
      doi = {10.48550/arXiv.2607.01918},
    }

15. ICML

    DropoutTS: Sample-Adaptive Dropout for Robust Time Series Forecasting

    Siru Zhong, Yiqiu Liu, Zhiqing Cui, Zezhi Shao, Fei Wang, Qingsong Wen, and Yuxuan Liang

    In The Forty-Third International Conference on Machine Learning, 2026

    摘要 DOI arXiv 引用 PDF 代码

    

    Deep time series models are vulnerable to noisy data ubiquitous in real-world applications. Existing robustness strategies either prune data or rely on costly prior quantification, failing to balance effectiveness and efficiency. In this paper, we introduce DropoutTS, a model-agnostic plugin that shifts the paradigm from "what" to learn to "how much" to learn. DropoutTS employs a Sample-Adaptive Dropout mechanism: leveraging spectral sparsity to efficiently quantify instance-level noise via reconstruction residuals, it dynamically calibrates model learning capacity by mapping noise to adaptive dropout rates - selectively suppressing spurious fluctuations while preserving fine-grained fidelity. Extensive experiments across diverse noise regimes and open benchmarks show DropoutTS consistently boosts superior backbones’ performance, delivering advanced robustness with negligible parameter overhead and no architectural modifications. Our code is available at this https URL.

    @inproceedings{zhong2026dropoutts,
      title = {DropoutTS: Sample-Adaptive Dropout for Robust Time Series Forecasting},
      author = {Zhong, Siru and Liu, Yiqiu and Cui, Zhiqing and Shao, Zezhi and Wang, Fei and Wen, Qingsong and Liang, Yuxuan},
      booktitle = {The Forty-Third International Conference on Machine Learning},
      year = {2026},
      url = {https://arxiv.org/abs/2601.21726},
      doi = {10.48550/arXiv.2601.21726},
    }

16. ICML

    PULSE: Generative Phase Evolution for Non-Stationary Time Series Forecasting

    Yangyou Liu, Zezhi Shao, Xinyu Chen, Hu Chen, Fei Wang, and Yuankai Wu

    In The Forty-Third International Conference on Machine Learning, 2026

    摘要 DOI arXiv 引用 PDF 代码

    

    Time series forecasting under non-stationarity faces a fundamental tension between capturing stable representations and adapting to distribution shifts. Existing methods implicitly rely on static historical assumptions, leading to a critical failure mode we term Phase Amnesia, where models become blind to the evolving global context. To resolve this, we formalize non-stationary dynamics through three physical hypotheses: wold decomposition, dynamical phase evolution, and heteroscedastic manifold generation. These principles inspire PULSE, a physics-informed, plug-and-play framework adopting a Disentangle–Evolve–Simulate design philosophy. Specifically, PULSE utilizes phase-anchored disentanglement to resolve optimization interference caused by dominant trends, employs a Phase Router to actively generate future trajectories, and introduces Statistic-Aware Mixup (SAM) to ensure robustness against out-of-distribution volatility. Empirically, PULSE enables a simple MLP backbone to achieve state-of-the-art or highly competitive performance across 12 real-world benchmarks. This validates that a correct physics-informed inductive bias is far more critical than raw architectural complexity for non-stationary forecasting. The code is available at: this https URL.

    @inproceedings{liu2026pulse,
      title = {PULSE: Generative Phase Evolution for Non-Stationary Time Series Forecasting},
      author = {Liu, Yangyou and Shao, Zezhi and Chen, Xinyu and Chen, Hu and Wang, Fei and Wu, Yuankai},
      booktitle = {The Forty-Third International Conference on Machine Learning},
      year = {2026},
      url = {https://arxiv.org/abs/2605.16793},
      doi = {10.48550/arXiv.2605.16793},
    }