Zexin Wang

Zexin Wang

AIOps · Time Series · AI Infra · AgentOps

Zexin Wang - Homepage

I am an Assistant Researcher and Postdoctoral Researcher at the Computer Network Information Center, Chinese Academy of Sciences, working with Prof. Gaogang Xie as my postdoctoral advisor. I received my Ph.D. in Computer Systems Architecture from University of Chinese Academy of Sciences and Computer Network Information Center, CAS in January 2026, advised by Prof. Jianhui Li and Assoc. Prof. Changhua Pei. Previously, I received my Bachelor’s degree in Computer Science and Technology from Peking University in 2020.

My research interests include AIOps, time series, AI infrastructure (AI Infra), and AgentOps. I am particularly interested in building reliable, interpretable, and practical intelligent operation methods for real-world systems, including anomaly detection, incident triage, root cause analysis, observability data modeling, AI cluster operations, and agent system operations.

I have published 10+ papers in top international conferences including KDD, WWW, ACL, and other venues. I have led or participated in industry-academia collaborations with Huawei, ZTE, and other partners, and have contributed to national key R&D and NDRC projects.

Feel free to drop me an email if you are interested in my research or just want to talk with me.

News

  • 2026.08: Four papers were accepted to ACM SIGKDD 2026: Rethinking Time Series Anomaly Detection from a Dynamic Perspective: Temporal-Frequency-Curvature Fusion; TSLoc: Self-Supervised Faulty Node Localization Framework in Large-scale Training Clusters; TSRBench: Benchmarking Time-Series Retrieval; and Don’t Predict, Prioritize: Rethinking GPU Reliability Assessment.
  • 2026.08: Smart Brain: Semantic Anomaly Detection for Operational Time Series in Large Scale Service Systems was accepted to the ASE 2026 Research Track.
  • 2026.08: Two papers were accepted to ISSRE 2026: BTSB: Business-Level Time Series Anomaly Detection Benchmark; and From Benchmark Accuracy to Diagnostic Boundaries: A Time-Aware Replication of DEST.
  • 2026.08: Title It Right: Faithful and Attractive Paper Title Generation via Pareto Dominance Alignment was accepted to EMNLP 2026.
  • 2026.08: Agent System Operations: Categorization, Challenges, and Future Directions was accepted by IEEE Transactions on Software Engineering (TSE).
  • 2026.02: Joined Computer Network Information Center, Chinese Academy of Sciences as an Assistant Researcher and Postdoctoral Researcher.
  • 2026: Received Beijing Outstanding Graduate and UCAS Outstanding Graduate awards.
  • 2025: Received the National Scholarship.
  • 2024: Received the CNIC Excellence Graduate Student Special Award.

Research Interests

  • AIOps: anomaly detection, incident triage, root cause analysis, and automated operations.
  • Time series: anomaly detection and unified reasoning over temporal data.
  • AI Infra: reliability, observability, and operations for AI infrastructure and large-scale training clusters.
  • AgentOps: observability, reliability, governance, and evaluation for agent systems.

Education and Experience

  • 2026.02 - Present, Assistant Researcher and Postdoctoral Researcher, Computer Network Information Center, Chinese Academy of Sciences, Computer Science and Technology. Postdoctoral advisor: Prof. Gaogang Xie.
  • 2020.09 - 2026.01, Ph.D., Computer Systems Architecture, University of Chinese Academy of Sciences / Computer Network Information Center, Chinese Academy of Sciences. Advisors: Prof. Jianhui Li and Assoc. Prof. Changhua Pei.
  • 2016.09 - 2020.07, B.S., Computer Science and Technology, School of Electronics Engineering and Computer Science, Peking University.

Selected Publications

  1. Ke Xiang, Yingchao Piao, Zexin Wang, Yufei Hou, Jianhui Li, Jingjing Li, Zihan Liu, Hang Cui, and Changhua Pei. Title It Right: Faithful and Attractive Paper Title Generation via Pareto Dominance Alignment. EMNLP 2026.
  2. Hang Cui, Cenjie Hu, Changhua Pei, Juncheng Hu, Haotian Si, Zihan Liu, Ke Xiang, Yuxuan Li, Quan Zhou, Xiaohui Nie, Zexin Wang, Jingjing Li, Dan Pei, and Gaogang Xie. From Benchmark Accuracy to Diagnostic Boundaries: A Time-Aware Replication of DEST. ISSRE 2026 RENE Track.
  3. Hang Cui, Cenjie Hu, Zexin Wang, Juncheng Hu, Haotian Si, Zihan Liu, Yuxuan Li, Jingjing Li, Dan Pei, Changhua Pei, and Gaogang Xie. BTSB: Business-Level Time Series Anomaly Detection Benchmark. ISSRE 2026 Research Track.
  4. Hang Cui, Zexin Wang, Jingjing Li, Juncheng Hu, Haotian Si, Cenjie Hu, Quan Zhou, Yongchang Hu, Lei Han, Dan Pei, Changhua Pei, and Gaogang Xie. Smart Brain: Semantic Anomaly Detection for Operational Time Series in Large Scale Service Systems. ASE 2026 Research Track.
  5. Hang Cui, Cenjie Hu, Zexin Wang, Jingjing Li, Juncheng Hu, Dan Pei, Changhua Pei, and Gaogang Xie. Predict Boldly, Recover Cautiously: Fast On-Router Route Anomaly Prediction and Recovery. ACM SIGCOMM 2026, pp. 2253-2255.
  6. Yunfei Zhang, Boyu Feng, Changhua Pei, Zexin Wang, Zhihuang Peng, Xinlong Liu, Hengyue Jiang, Difeng Ma, Jiayi Zhang, Yongzhou Yao, Yanan Zhao, Fei Sun, Yintong Huo, Zhaoyang Liu, Jingjing Li, Gaogang Xie, and Dan Pei. LongRCA Bench: Diagnosing Responsible Roles and Root Causes in Long-Horizon Agent Failures. arXiv, 2026.
  7. Difeng Ma, Changhua Pei, Yuanwei Lu, Quan Zhou, Zexin Wang, Yibo Zhu, Daxin Jiang, Dan Pei, Jingjing Li, and Gaogang Xie. Don't Predict, Prioritize: Rethinking GPU Reliability Assessment. KDD 2026 Applied Data Science Track.
  8. Cenjie Hu, Hang Cui, Zexin Wang, Juncheng Bao, Jingwen Yang, Jingjing Li, Changhua Pei, Dan Pei, and Gaogang Xie. TSRBench: Benchmarking Time-Series Retrieval. KDD 2026 Datasets and Benchmarks Track.
  9. Quan Zhou, Changhua Pei, Yuanwei Lu, Difeng Ma, Zexin Wang, Jianhui Li, Yibo Zhu, Daxin Jiang, Dan Pei, Jingjing Li, and Gaogang Xie. TSLoc: Self-Supervised Faulty Node Localization Framework in Large-scale Training Clusters. KDD 2026 Applied Data Science Track.
  10. Hang Cui, Zexin Wang, Changhua Pei, Juncheng Hu, Haotian Si, Quan Zhou, Cenjie Hu, Jingjing Li, Dan Pei, and Gaogang Xie. Rethinking Time Series Anomaly Detection from a Dynamic Perspective: Temporal-Frequency-Curvature Fusion. KDD 2026 Research Track.
  11. Zihan Liu, Jianhui Li, Zexin Wang, Fei Sun, Jingjing Li, Zheyuan Li, Ke Xiang, Hang Cui, Houhua Gong, Changhua Pei, and Gaogang Xie. EviReport: From Reasoned Outlines to Evidence Tracked Long-Form Reports. Findings of ACL 2026.
  12. Changhua Pei, Zheyuan Li, Zexin Wang, Hang Cui, Xiaohui Nie, Qi Zhou, Fang Situ, Cheng Zhang, Xin Zhang, Xidao Wen, Gaogang Xie, Jingjing Li, and Dan Pei. UModel: An Agent-Ready Observability Data Modeling Method at Scale. arXiv, 2026.
  13. Zexin Wang, Changhua Pei, Yuanhao Liu, Jingjing Li, Yintong Huo, Quan Zhou, Haotian Si, Hang Cui, Zihan Liu, Jianhui Li, Gaogang Xie, Fei Sun, Dan Pei, and David Lo. Agent System Operations: Categorization, Challenges, and Future Directions. IEEE Transactions on Software Engineering, 2026.
  14. Changhua Pei, Hang Cui, Jingjing Li, Yuxuan Li, Zihan Liu, Xinyuan Liao, Cenjie Hu, Jiabao Wang, Zheyuan Li, Zexin Wang, Haotian Si, Ke Xiang, Gaogang Xie, and Dan Pei. Smart Eye: LLM-Guided Proposer-Verifier Framework for Industrial-Scale Log Anomaly Detection. WWW 2026.
  15. Zexin Wang, Changhua Pei, Yang Liu, Hengyue Jiang, Quan Zhou, Haotian Si, Hang Cui, Jianhui Li, Gaogang Xie, Jingjing Li, and Dan Pei. ViTs: Teaching Machines to See Time Series Anomalies Like Human Experts. WWW 2026.
  16. Haotian Si, Changhua Pei, Xiao He, Zeyan Li, Zhe Xie, Zexin Wang, Jiyao Hu, Zhaoyang Yu, Tieying Zhang, Dan Pei, Jianhui Li, and Gaogang Xie. KairosVL: Orchestrating Time Series and Semantics for Unified Reasoning. arXiv, 2026.
  17. Changhua Pei, Zexin Wang, Fengrui Liu, Zeyan Li, Yang Liu, Xiao He, Rong Kang, Tieying Zhang, Jianjun Chen, Jianhui Li, Gaogang Xie, and Dan Pei. Flow-of-Action: SOP Enhanced LLM-Based Multi-Agent System for Root Cause Analysis. WWW Companion 2025.
  18. Yang Liu, Yonghua Zhao, Zexin Wang, Rongfeng Huang, Dingye Zhang, and Xinyin Zhang. Efficient Implementation of the LOBPCG Algorithm on a CPU-GPU Cluster. NPC 2024, pp. 65-76.
  19. Yue Wang, Changhua Pei, Zexin Wang, Yingqiang Wang, Guo Chen, Yuchao Zhang, Yi Li, Jingjing Li, Jianhui Li, and Gaogang Xie. ActiveDNS: Is There Room for DNS Optimization Beyond CDNs? LCN 2024.
  20. Zexin Wang, Jianhui Li, Minghua Ma, Ze Li, Yu Kang, Chaoyun Zhang, Chetan Bansal, Murali Chintalapati, Saravan Rajmohan, Qingwei Lin, Dongmei Zhang, Changhua Pei, and Gaogang Xie. Large Language Models Can Provide Accurate and Interpretable Incident Triage. ISSRE 2024.
  21. Zexin Wang, Changhua Pei, Minghua Ma, Xin Wang, Zhihan Li, Dan Pei, Saravan Rajmohan, Dongmei Zhang, Qingwei Lin, Haiming Zhang, Jianhui Li, and Gaogang Xie. Revisiting VAE for Unsupervised Time Series Anomaly Detection: A Frequency Perspective. WWW 2024.

Invited Talks

  • IWQoS 2026 Workshop, Invited Keynote: AI cluster operations experience.
  • Beautiful China 2026 Digital Governance Conference, Invited Talk: SDG Lingxi foundation model.