Xi Jiang — Academic Homepage

About

I am a Ph.D. candidate in Computer Science at the Southern University of Science and Technology (SUSTech), advised by Prof. Feng Zheng.

My research focuses on industrial anomaly detection, multimodal large language models, and agentic systems. I aim to build inspection systems that can compare visual evidence, explain anomalies, and actively acquire useful observations. My work has received 550+ citations, and the Awesome Industrial Anomaly Detection repository I maintain has 4k+ stars.

I received my M.Sc. in Computer Science from SUSTech in 2023 and my B.Sc. in Computer Science from Xi’an Jiaotong University in 2020, where I was advised by Prof. Danfeng Shan. I was awarded the National Scholarship for Doctoral Students in 2025.

From Oct. 2025 to Apr. 2026, I was a joint-training Ph.D. student at A*STAR IHPC–CFAR in Singapore, supervised by Joey Tianyi Zhou. Before that, I spent nearly four years as a research intern at Tencent YouTu Lab (2022–2025), and earlier interned at Huawei on continual and federated learning.

News

2026.06: Released our survey on recent advances in industrial anomaly detection.
2026.05: AnomalyClaw preprint and code were released.
2026.03: AD-Copilot preprint and code were released.
2025.05: SoftPatch+ appeared in Pattern Recognition.
2025.01: MMAD was accepted to ICLR 2025.
2025: Received the National Scholarship for Doctoral Students.

Earlier news

2025: Received the SUSTech Special Scholarship.
2024: Received an Honorable Mention in the CVPR VAND 2.0 Challenge; our image customization work was accepted to ECCV 2024.
2023.09: Our Bayesian federated learning work was accepted by IEEE TPAMI.
2022: SoftPatch was accepted to NeurIPS 2022.

Publications

View all

^* Equal contribution. Unaccepted work is listed only when I am first or co-first author.

Overview of the MMAD benchmark and evaluation dimensions

ICLR 2025 Paper Code

MMAD: The Comprehensive Benchmark for Multimodal Large Language Models in Industrial Anomaly Detection

Xi Jiang, Jian Li, Hanqiu Deng, Yong Liu, Bin-Bin Gao, Yifeng Zhou, Jialin Li, Chengjie Wang, Feng Zheng

A comprehensive benchmark covering anomaly discrimination, localization, description, object analysis, and defect reasoning.

Framework of SoftPatch+ for noise discovery and coreset construction

Pattern Recognition 2025 Paper Code

SoftPatch+: Fully Unsupervised Anomaly Classification and Segmentation

Chengjie Wang^*, Xi Jiang^*, Bin-Bin Gao, Zhenye Gan, Yong Liu, Feng Zheng, Lizhuang Ma

Extends patch-level denoising with multiple discriminators for robust fully unsupervised inspection under high noise ratios.

Tuning-free image customization pipeline with image and text guidance

ECCV 2024Paper Project

Tuning-Free Image Customization with Image and Text Guidance

Pengzhi Li, Qiang Nie, Ying Chen, Xi Jiang, Kai Wu, Yuhuan Lin, Yong Liu, Jinlong Peng, Chengjie Wang, Feng Zheng

A tuning-free diffusion framework for precise local customization jointly guided by reference images and text.

Server-side and client-side overview of the Bayesian federated learning framework

IEEE TPAMI 2024 Paper Code

A Bayesian Federated Learning Framework with Online Laplace Approximation

Liangxi Liu^*, Xi Jiang^*, Feng Zheng, Hong Chen, Guo-Jun Qi, Heng Huang, Ling Shao

Approximates client and server posteriors online to reduce aggregation error and local forgetting under heterogeneous data.

Intuition of SoftPatch for learning from noisy training data

NeurIPS 2022 Paper Code

SoftPatch: Unsupervised Anomaly Detection with Noisy Data

Xi Jiang, Jianlin Liu, Jinbao Wang, Qiang Nie, Kai Wu, Yong Liu, Chengjie Wang, Feng Zheng

A patch-level noise discriminator and soft-weighted coreset that make normal-only anomaly detection robust to contaminated training data.

Architecture of the AD-Copilot comparison encoder

Preprint · 2026 Paper Code

AD-Copilot: A Vision-Language Assistant for Industrial Anomaly Detection via Visual In-context Comparison

Xi Jiang, Yue Guo, Jian Li, Yong Liu, Bin-Bin Gao, Hanqiu Deng, Jun Liu, Heng Zhao, Chengjie Wang, Feng Zheng

A comparison-aware multimodal assistant that learns fine-grained anomaly perception and explanation from paired visual evidence.

2026

Embodied-AD: Industrial Anomaly Detection through Robotic View Acquisition

Xi Jiang, Zesheng Yang, Bingzhang Hu, Feng Zheng

Frames inspection as active robotic perception, using normal 3D Gaussian templates for pose-aligned reference rendering and view selection.

AnomalyClaw tool-grounded refutation framework

Preprint · 2026 Paper Code

AnomalyClaw: A Universal Visual Anomaly Detection Agent via Tool-Grounded Refutation

Xi Jiang, Yinjie Zhao, Zesheng Yang, Feng Zheng

A training-free anomaly agent that challenges its first judgment through multi-turn, tool-grounded refutation.

Preprint · 2026Paper

A Survey of Recent Advances in Industrial Anomaly Detection: From Normal-Only Training to Foundation-Model Priors

Xi Jiang, Bingzhang Hu, Feng Zheng

Organizes recent progress into five foundation-model-driven generalization mechanisms and outlines open challenges.

UniPCB data and vision-language benchmark generation framework

Preprint · 2026Paper

UniPCB: A Unified Vision-Language Benchmark for Open-Ended PCB Quality Inspection

Fuxiang Sun^*, Xi Jiang^*, Jiansheng Wu, Haigang Zhang, Feng Zheng, Jinfeng Yang

A unified PCB inspection benchmark and progressive curriculum for domain-specialized multimodal reasoning.

Intuition for negation-aware visual grounding

Preprint · 2026Paper

Mastering Negation: Boosting Grounding Models via Grouped Opposition-Based Learning

Zesheng Yang^*, Xi Jiang^*, Bingzhang Hu, Weili Guan, Runmin Cong, Guo-Jun Qi, Feng Zheng

Introduces positive–negative semantic groups and opposition-based objectives for negation-aware visual grounding.

MINT-AD class-aware unified anomaly detection framework

Preprint · 2024Paper

Toward Multi-class Anomaly Detection: Exploring Class-aware Unified Model against Inter-class Interference

Xi Jiang, Ying Chen, Qiang Nie, Jianlin Liu, Yong Liu, Chengjie Wang, Feng Zheng

MINT-AD learns class-aware query embeddings to reduce inter-class interference in a unified anomaly detector.

Experience & education

Experience

2025.10–2026.04 A*STAR, Singapore Joint-training Ph.D. · IHPC–CFAR · Advisor: Joey Tianyi Zhou
2022.01–2025.10 Tencent YouTu Lab Research Intern · Industrial AI · CSIG Fuxi Research Center
2021.09–2022.01 Huawei Research Intern · Continual & federated learning

Education

2023–present Ph.D. in Computer Science SUSTech · Advisor: Prof. Feng Zheng
2020–2023 M.Sc. in Computer Science SUSTech
2016–2020 B.Sc. in Computer Science Xi’an Jiaotong University · Advisor: Prof. Danfeng Shan

Honors & service

2025National Scholarship for Doctoral Students
2025SUSTech Special Scholarship
2024CVPR VAND 2.0 Challenge — Honorable Mention
2023–2025Outstanding Graduate Student Award
2017XJTU Siyuan Scholarship; 2nd Prize, 19th China Robot Competition
ServiceReviewer for ICLR, NeurIPS, CVPR, ICCV, ECCV, CoRL, AAAI, TMM, ACM CSUR, Pattern Recognition, and more

For research collaboration or open-source work, email jiangx2020@mail.sustech.edu.cn.