About Me

Hi! I’m Shilong Bao (包世龙, E-mail: baoshilong@ict.ac.cn). I am currently an Assistant Research Fellow at the Institute of Computing Technology, Chinese Academy of Sciences (ICT, CAS). I received my Ph.D. degree from the Institute of Information Engineering, Chinese Academy of Sciences (IIE, CAS), supervised by Prof. Qingming Huang (黄庆明) (IEEE Fellow). I am also fortunate to collaborate with Qianqian Xu (许倩倩) (Professor at ICT, CAS), Xiaochun Cao (操晓春) (Dean of the School of Cyber Science and Technology, Sun Yat-sen University), and Zhiyong Yang (杨智勇) (Tenure-track Assistant Professor at UCAS).

My research primarily focuses on machine learning and AI safety, with particular interests in:

Trustworthy Machine Learning (Robust Learning, Imbalanced Learning, and Ranking & AUC Optimization)
Safe and Reliable Generative AI (Safety, Fairness, Robustness, and Copyright Protection)

欢迎加入科研实习（可远程）

招募本科生和研究生，开展可信机器学习与安全可靠生成式人工智能研究，欢迎通过邮件联系。

欢迎学术界与工业界交流合作

欢迎学术界与工业界围绕相关方向交流合作，可通过邮件联系。

🔥 News

2026.08.05: 🎉🎉 I chaired the Forum on Efficient Training and Inference of Large Models and gave an invited talk, “A Brief Discussion on Representation Reconcilement Learning in Multimodal Models”, at the CSIG Young Scientists Conference 2026.
2026.08.05: 🎉🎉 I am honored to serve as an Area Chair for GroundLM 2026, an EMNLP 2026 Workshop.
2026.07.16: 🎉🎉 I joined the Institute of Computing Technology, Chinese Academy of Sciences (ICT, CAS) as an Assistant Research Fellow.
2026.06.03: 🎉🎉 Our team won the 1st Place Award in the CVPR 2026 Vision-based Assistants in the Real World Workshop, AI Coach Challenge (Cooking Track).
2026.06.03: 🎉🎉 Our team won the 1st Place Award in CVPR EgoVis HoloAssist Challenges for Fine-grained Video Understanding (Mistake Detection Track, 2026), successfully defending our 2025 title in the same track.
2026.05.14: 🎉🎉 I have been recognized as an ICML 2026 Gold Reviewer.
2026.04.28: 🎉🎉 One paper has been accepted by ICML 2026 with Oral presentation (0.69%). Congratulations to Shixi!
2026.02.21: 🎉🎉 Two papers have been accepted by CVPR 2026. Congratulations to Boyu and Feiran!
2025.11.15: One paper has been accepted by T-PAMI 2025!
2025.10.12: 🎉🎉 My PhD Thesis “Toward Efficient and Generalizable Collaborative Metric Learning Algorithms” (in Chinese) has been selected as the ACM China Excellent Doctoral Dissertation Award Nomination (totally 5 papers in China) (ACM China 优博奖提名)
2025.09.20: 🎉🎉 One paper has been accepted by NeurIPS 2025!
2025.09.15: 🎉🎉 One paper has been accepted by T-PAMI!
2025.09.13: 🎉🎉 We are organizing the forum “Efficient Training and Inference of Large Models” at the CSIG Young Scientists Conference 2025. Welcome to join us!
2025.08.01: 🎉🎉 Our team won the 1st Place Award in ICCV 2025 Competition for High-Quality Face Dataset Generation (DataCV Challenge), with one paper accepted by ICCV 2025 workshop!
2025.06.30: 🎉🎉 My PhD Thesis “Toward Efficient and Generalizable Collaborative Metric Learning Algorithms” (in Chinese) has been selected as the Distinguished Dissertation Award of Chinese Academy of Sciences (totally 100 papers) (中国科学院百篇优博论文)
2025.06.18: 🎉🎉 Our team (MR-CAS) won the 1st Place Award in CVPR 2025 Workshop on Compositional 3D Vision (C3DV 3DCoMPaT-200, Coarse-Grained GCR Track)
2025.06.12: 🎉🎉 Our team (MR-CAS) won the 1st Place Award in CVPR 2025 Competition for Fine-grained Video Understanding (EgoVis HoloAssist Challenges: Mistake Detection Track).
2025.05.20: 🎉🎉 I have been nominated as ICLR Notable Reviewer 2025.
2025.05.02: 🎉🎉 Three papers have been accepted by ICML 2025.
2025.02.20: 🎉🎉 One paper has been accepted by T-PAMI 2025.

✨ Highlight Papers

T-PAMI 2026

Towards Size-invariant Salient Object Detection: A Generic Evaluation and Optimization Approach.

Shilong Bao, Qianqian Xu, Feiran Li, Boyu Han, Zhiyong Yang, Xiaochun Cao, and Qingming Huang. IEEE Transactions on Pattern Analysis and Machine Intelligence (T-PAMI), 48(2): 1115–1132, Feb. 2026.

T-PAMI 2025

AUCPro: AUC-Oriented Provable Robustness Learning.

Shilong Bao, Qianqian Xu, Zhiyong Yang, Yuan He, Xiaochun Cao, and Qingming Huang. IEEE Transactions on Pattern Analysis and Machine Intelligence (T-PAMI), 47(6): 4579-4596, Jun. 2025. |[Code]|

T-PAMI 2024

Improved Diversity-Promoting Collaborative Metric Learning for Recommendation.

Shilong Bao, Qianqian Xu, Zhiyong Yang, Yuan He, Xiaochun Cao, and Qingming Huang. IEEE Transactions on Pattern Analysis and Machine Intelligence (T-PAMI), 46(12): 9004–9022, Dec. 2024.

T-PAMI 2023

Rethinking Collaborative Metric Learning: Toward an Efficient Alternative without Negative Sampling.

Shilong Bao, Qianqian Xu, Zhiyong Yang, Xiaochun Cao and Qingming Huang. IEEE Transactions on Pattern Analysis and Machine Intelligence (T-PAMI), 45(1): 1017-1035, Jan. 2023. |[Code]|

NeurIPS 2022 (Oral)

The Minority Matters: A Diversity-Promoting Collaborative Metric Learning Algorithm.

Shilong Bao, Qianqian Xu, Zhiyong Yang, Yuan He, Xiaochun Cao, and Qingming Huang. Advances in Neural Information Processing Systems (NeurIPS), 35: 2451–2464, 2022. (Oral, 1.7%) | [Code]| [Video] | [Poster] | [Slides]

ICML 2026 (Oral)

GoodDiffusion: Proactive Copyright Protection for Diffusion Generative Models via Learnable Sample-specific Signatures.

Shixi Qin, Zhiyong Yang, Shilong Bao, Zitai Wang, Qianqian Xu, and Qingming Huang. International Conference on Machine Learning (ICML), 2026. (Oral, 0.69%) | [Code]|

ICML 2021 (Long Talk)

When All We Need is a Piece of the Pie: A Generic Framework for Optimizing Two-way Partial AUC.

Zhiyong Yang, Qianqian Xu, Shilong Bao, Yuan He, Xiaochun Cao, and Qingming Huang. International Conference on Machine Learning (ICML), PMLR 139: 11820–11829, 2021. (Long Talk, 3%) | [Code]| [Video] | [Poster] | [Slides]

ICML 2024 (Spotlight)

Size-invariance Matters: Rethinking Metrics and Losses for Imbalanced Multi-object Salient Object Detection.

Feiran Li, Qianqian Xu, Shilong Bao, Zhiyong Yang, Runmin Cong, Xiaochun Cao, and Qingming Huang. International Conference on Machine Learning (ICML), PMLR 235: 28989–29021, 2024. (Spotlight, 3.5%) | [Code]| [Video] | [Poster] | [Slides] | [Project]

📝 Publications

2026

Guiding Diffusion-based Reconstruction with Contrastive Signals for Balanced Visual Representation Boyu Han, Qianqian Xu, Shilong Bao, Zhiyong Yang, Ruochen Cui, Xilin Zhao, and Qingming Huang. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2369–2380, 2026. |[Code]|
BlackMirror: Black-Box Backdoor Detection for Text-to-Image Models via Instruction-Response Deviation Feiran Li, Qianqian Xu, Shilong Bao, Zhiyong Yang, Xilin Zhao, Xiaochun Cao, and Qingming Huang. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 30098–30109, 2026. |[Code]|
Closing the Approximation Gap of Partial AUC Optimization: A Tale of Two Formulations Yangbangyan Jiang, Qianqian Xu, Huiyang Shao, Zhiyong Yang, Shilong Bao, Xiaochun Cao, and Qingming Huang. IEEE Transactions on Pattern Analysis and Machine Intelligence (T-PAMI), 48(3): 3482–3498, Mar. 2026.

2025

LightFair: Towards an Efficient Alternative for Fair T2I Diffusion via Debiasing Pre-trained Text Encoders Boyu Han, Qianqian Xu, Shilong Bao, Zhiyong Yang, Kangli Zi, and Qingming Huang. Advances in Neural Information Processing Systems (NeurIPS), 38, 2025. |[Code]|
One Image is Worth a Thousand Words: A Usability Preservable Text-Image Collaborative Erasing Framework Feiran Li, Qianqian Xu, Shilong Bao, Zhiyong Yang, Xiaochun Cao, and Qingming Huang. International Conference on Machine Learning (ICML), PMLR 267: 36486–36529, 2025. |[Code]|
MixBridge: Heterogeneous Image-to-Image Backdoor Attack through Mixture of Schrödinger Bridges Shixi Qin, Zhiyong Yang, Shilong Bao, Shi Wang, Qianqian Xu, and Qingming Huang. International Conference on Machine Learning (ICML), PMLR 267: 50397–50434, 2025. |[Code]|
OpenworldAUC: Towards Unified Evaluation and Optimization for Open-world Prompt Tuning Cong Hua, Qianqian Xu, Zhiyong Yang, Zitai Wang, Shilong Bao, and Qingming Huang. International Conference on Machine Learning (ICML), PMLR 267: 24975–25020, 2025. |[Code]|
Bidirectional Logits Tree: Pursuing Granularity Reconcilement in Fine-Grained Classification Zhiguang Lu, Qianqian Xu, Shilong Bao, Zhiyong Yang, and Qingming Huang. AAAI Conference on Artificial Intelligence (AAAI), 39(18): 19189–19197, 2025. |[Code]|

2024

AUCSeg: AUC-oriented Pixel-level Long-tail Semantic Segmentation Boyu Han, Qianqian Xu, Zhiyong Yang, Shilong Bao, Peisong Wen, Yangbangyan Jiang, and Qingming Huang. Advances in Neural Information Processing Systems (NeurIPS), 37: 126863–126907, 2024. |[Code]|
ReconBoost: Boosting Can Achieve Modality Reconcilement Cong Hua, Qianqian Xu, Shilong Bao, Zhiyong Yang, and Qingming Huang. International Conference on Machine Learning (ICML), PMLR 235: 19573–19597, 2024. | [Code] |
Harnessing Hierarchical Label Distribution Variations in Test Agnostic Long-tail Recognition Zhiyong Yang, Qianqian Xu, Zitai Wang, Sicong Li, Boyu Han, Shilong Bao, Xiaochun Cao, and Qingming Huang. International Conference on Machine Learning (ICML), PMLR 235: 56624–56664, 2024. | [Code] |

Earlier Publications

Revisiting AUC-oriented Adversarial Training with Loss-Agnostic Perturbations Zhiyong Yang, Qianqian Xu, Wenzheng Hou, Shilong Bao, Yuan He, Xiaochun Cao, and Qingming Huang. IEEE Transactions on Pattern Analysis and Machine Intelligence (T-PAMI), 45(12): 15494–15511, Dec. 2023. | [Code] |
AUC-Oriented Domain Adaptation: From Theory to Algorithm Zhiyong Yang, Qianqian Xu, Shilong Bao, Peisong Wen, Yuan He, Xiaochun Cao, and Qingming Huang. IEEE Transactions on Pattern Analysis and Machine Intelligence (T-PAMI), 45(12): 14161–14174, Dec. 2023. | [Code] |
Optimizing Two-way Partial AUC with an End-to-end Framework Zhiyong Yang, Qianqian Xu, Shilong Bao, Yuan He, Xiaochun Cao, and Qingming Huang. IEEE Transactions on Pattern Analysis and Machine Intelligence (T-PAMI), 45(8): 10228–10246, Aug. 2023. | [Code] |
Asymptotically Unbiased Instance-wise Regularized Partial AUC Optimization: Theory and Algorithm Huiyang Shao, Qianqian Xu, Zhiyong Yang, Shilong Bao, and Qingming Huang. Advances in Neural Information Processing Systems (NeurIPS), 35: 38667–38679, 2022. | [Code] |
AdAUC: End-to-end Adversarial AUC Optimization Against Long-tail Problems Wenzheng Hou, Qianqian Xu, Zhiyong Yang, Shilong Bao, Yuan He, and Qingming Huang. International Conference on Machine Learning (ICML), PMLR 162: 8903–8925, 2022. | [Code] |
Learning with Multiclass AUC: Theory and Algorithms Zhiyong Yang, Qianqian Xu, Shilong Bao, Xiaochun Cao, and Qingming Huang. IEEE Transactions on Pattern Analysis and Machine Intelligence (T-PAMI), 44(11): 7747–7763, Nov. 2022. | [Code] |
Collaborative Preference Embedding against Sparse Labels Shilong Bao, Qianqian Xu, Ke Ma, Zhiyong Yang, Xiaochun Cao, and Qingming Huang. ACM International Conference on Multimedia (ACM-MM), 2079–2087, 2019. | [Code]|

📖 Services

Conferences

GroundLM 2026 (EMNLP 2026 Workshop): Area Chair
ICML: PC Member (2022-2026)
ICLR: PC Member (2024-2026)
NeurIPS: PC Member (2023-2026)
CVPR: PC Member (2024-2026)
ICCV: PC Member (2025)
WACV: PC Member (2025)
AAAI: PC Member (2023-2026)
AISTATS: PC Member (2025-2026)

Journals

IEEE Transactions on Pattern Analysis and Machine Intelligence (T-PAMI): Reviewer
IEEE Transactions on Multimedia (T-MM): Reviewer
IEEE Transactions on Circuits and Systems for Video Technology (T-CSVT): Reviewer
ACM Transactions on Multimedia Computing, Communications and Applications (TOMM): Reviewer
Multimedia Systems: Reviewer

Others

2026.08.05 Chair of the Forum on Efficient Training and Inference of Large Models at the CSIG Young Scientists Conference 2026
2026.07 Forum Co-chair of the Theoretical Foundations of Trustworthy Artificial Intelligence for Multimedia Forum at ChinaMM 2026
2026.05 Co-chair of Forum on Trustworthy Multimedia Analysis and Privacy-Preserving Computing at the CCIG 2026
2025.11 Program Chair of Beijing Youth Science and Technology Salon: Multimodal Intelligent Perception and Cross-modal Computing (北京青年科技沙龙)
2025.08 Co-chair of Efficient Training and Inference of Large Models at the CSIG Young Scientists Conference 2025

🎖 Honors and Awards

2026 1st Place Award in CVPR 2026 Vision-based Assistants in the Real World Workshop (AI Coach Challenge, Cooking Track)
2026 1st Place Award in CVPR EgoVis HoloAssist Challenges for Fine-grained Video Understanding (Mistake Detection Track, successfully defending our 2025 title)
2026 ICML Gold Reviewer (ICML 2026)
2025 ACM China Excellent Doctoral Dissertation Award Nomination (ACM中国优博奖提名, 5 papers in China)
2025 ACM China SIGMM Excellent Doctoral Dissertation Award (ACM中国SigMM优博, 3 papers in total)
2025 Distinguished Dissertation Award of Chinese Academy of Sciences (totally 100 papers) (中国科学院优秀博士学位论文，中科院全学科100篇)
2025 1st Place Award in ICCV 2025 Competition for High-Quality Face Dataset Generation (DataCV Challenge)
2025 1st Place Award at the 3rd CVPR Workshop on Compositional 3D Vision (Coarse-Grained GCR Track Challenge, 2025)
2025 1st Place Award in CVPR EgoVis HoloAssist Challenges for Fine-grained Video Understanding (Mistake Detection Track, 2025)
2025 ICLR Notable Reviewer (480/all)
2025 Young Elite Scientists Sponsorship Program of the Beijing High Innovation Plan (北京”高创计划”-青年人才托举工程)
2024 Outstanding Doctoral Dissertation Award of Beijing Society of Image and Graphics (BSIG). (北京图象图形学学会优秀博士学位论文奖 ( 京津冀5篇 ))
2023 Zhuliyuehua Scholarship for Excellent Doctoral Student, CAS. (中国科学院朱李月华奖学金，中科院共300人)
2022 National Scholarship, Ministry of Education of the People’s Republic of China. (国家奖学金)
2021 Director Special Scholarship Award, IIE, CAS. (中科院信息工程研究所所长特别奖)
2017 The ACM-ICPC Asia Regional Contest Qingdao Site 2017 Silver Medal (ACM-ICPC 亚洲区域赛 (青岛站))
2017 The ACM-ICPC Asia Regional Contest Xian Site 2017 Bronze Medal (ACM-ICPC 亚洲区域赛 (西安站))
2017 3rd China Collegiate Programming Contest Harbin Site Bronze Medal (第三届中国大学生程序设计竞赛 CCPC (哈尔滨站))

🎓 Educations & Work Experience

2026.07 - Present, Assistant Research Fellow.

Institute of Computing Technology, Chinese Academy of Sciences (ICT, CAS), Beijing.

2024.07 - 2026.06, Postdoctoral Fellow.

School of Computer Science and Technology.

University of Chinese Academy of Sciences (UCAS), Beijing.

2019.09 - 2024.06, Ph.D. in Computer Applied Technology.

Institute of Information Engineering, Chinese Academy of Sciences (IIE, CAS).

University of Chinese Academy of Sciences, Beijing.

2015.09 - 2019.06, Undergraduate.

College of Computer Science and Technology.

Qingdao University (QDU), Qingdao.

💬 Invited Talks

Official visual for the Academic Rising Stars Forum at the CSIG Young Scientists Conference 2026

2026.08.05: A Brief Discussion on Representation Reconcilement Learning in Multimodal Models. [Website]

Abstract: Large-scale pretraining gives visual, language, and audio encoders strong general-purpose representation capabilities, supporting multimodal fusion, visual representation enhancement, and conditional generation. Yet differences among pretraining objectives, representation granularity, and downstream requirements create adaptation challenges when these representations are transferred, combined, or reused. Centered on multimodal representation reconcilement learning, this talk examines how conflicts among representation signals can be identified and mitigated through optimization. It discusses modality competition in multimodal fusion, objective conflicts in pretrained visual representations, and the effect of text-encoder bias on generative fairness, showing why effective reuse requires downstream-oriented representation reconcilement and adaptation.

MLNLP 2025 · Large Model Safety

2025.11: Towards Harmless Multimodal Generation: Challenges and Preliminary Pathways . [Website] | [Video]

Abstract: Generative AI is reshaping digital content creation, but its capacity to produce harmful content remains a major obstacle to real-world deployment. This talk presents our preliminary exploration of harmless generation along three directions. Targeted model unlearning is used to reduce harmful outputs, lightweight fairness interventions are introduced to mitigate generation bias, and analyses of backdoor vulnerabilities provide evidence for more robust defenses. Experimental results show that these approaches achieve encouraging effects while largely preserving model utility, offering concrete directions for further research on safer multimodal generation.

ChinaMM 2025 Special Forum 22 · Trustworthy Multimedia Learning

2025.08: Efficient, Generalizable, and Robust Collaborative Ranking Learning. [Website]

Abstract: Collaborative ranking is a fundamental technique supporting representation learning, content retrieval, and multimedia recommendation. When applied to large-scale, low-value-density, and highly diverse web data, it often faces limited representational capacity, low computational efficiency, and inadequate robustness, which constrain model generalization. Existing studies mainly focus on model architecture design and accelerated loss optimization, while systematic theoretical analysis of generalization remains limited. This talk introduces a theoretical framework for analyzing the generalization of collaborative ranking and uses the resulting theory to guide the design and optimization of ranking algorithms toward efficient, robust, and generalizable collaborative ranking.

2023.02: AI TIME Youth PhD Talk, NeurIPS 2022 session. [Video].

2022.11: Oral presentation at NeurIPS 2022. [Video].

💻 Fundings and Project

2025.08: Young Scientists Fund of the National Natural Science Foundation of China (NSFC青年基金C类, No.62502496, PI )
2025.07: General Program of the Chinese Postdoctoral Science Foundation (中国博士后科学基金面上资助, No.2025M771492, PI )
2025.06: CAS Special Research Assistant Talent Support Program (中国科学院特别研究助理资助项目, PI )
2025.07: Beijing Youth Science and Technology Salon (北京青年科技沙龙项目, PI )
2025.01: National Natural Science Foundation of China (NSFC), Special Project (NSFC专项项目, No. 62441232, Core Member )
2024.07: Postdoctoral Fellowship Program of the Chinese Postdoctoral Science Foundation (中国博士后科学基金会国家资助博士后研究人员计划（B档）, No.GZB20240729, PI )

XCurve

2020.02 - now: As a core member, I participated in the development of XCurve: Machine Learning with Decision-Invariant Metrics.

Machine learning and deep learning technologies have recently been successfully employed in many complicated high-stake decision-making applications. The goal of Xcurve learning is to learn high-quality models that can adapt to different decision conditions, which provides a systematic solution to optimize the area under different kinds of performance curves. Welcome to try now and give us feedback!