Hi, I am Zhiheng Fu (付志恒).
Welcome to my homepage! I am currently an undergraduate student in the School of Software, Shandong University, under the supervision of Prof. Liqiang Nie and Prof. Yupeng Hu, working closely with Dr. Zixu Li. My research interests mainly focus on Multimodal Large Language Models, robust representation learning, and trustworthy AI.
On the model side, I focus fine-grained visual-textual semantic fusion, composed image/video understanding, robust noise mitigation, knowledge calibration, and intent-aware representation learning, with representative works including Air-Know, TEMA, ConeSep, INTENT, FineCIR, and ENCODER. On the evaluation side, I have contributed to long-form video and egocentric vision reasoning systems, including the CVPRW challenge technical reports R3, TempRet, EgoAdapt, OmniEgo-R2, and EgoAction.
Beyond publications, I actively participate in open-source research and challenge-driven system building. I have led or contributed to projects accepted by ACL, CVPR, AAAI, and related workshops, and I have been a core member of teams that won champion, runner-up, and third-place awards in multiple CVPR 2026 challenges. My work has also been recognized by the National Scholarship and the Grand Prize in the CICAS Smart Power Scenario Competition.
欢迎来到我的主页！我目前是山东大学软件学院本科生，师从 Prof. Liqiang Nie 与 Prof. Yupeng Hu，并与 Dr. Zixu Li 保持紧密合作。我的研究兴趣主要聚焦于多模态大语言模型、鲁棒表征学习与可信AI研究。
一方面，我关注细粒度视觉-语言语义融合、组合式图文/视频理解、鲁棒噪声抑制、知识校准与意图感知表征学习，代表工作包括 Air-Know、TEMA、ConeSep、INTENT、FineCIR 与 ENCODER；另一方面，我参与长视频理解、第一视角视觉推理与多模态评测系统构建，相关技术报告包括 R3、TempRet、EgoAdapt、OmniEgo-R2 与 EgoAction。
除学术论文外，我也积极参与开源研究与挑战赛系统建设。相关工作发表于 ACL、CVPR、AAAI 及其 workshop，并作为核心成员参与团队在 CVPR 2026 多个国际挑战赛中获得冠军、亚军和季军。相关经历曾获国家奖学金与 CICAS 智能电力场景竞赛特等奖等荣誉。
I am a member of the Intelligent Media Research Center (iLearn). I believe open-source research makes multimodal learning more reproducible and collaborative. My major projects and our lab projects are open-source; please visit iLearn Lab and feel free to share your valuable feedback.
我隶属于智能媒体研究中心 (iLearn)，并相信开源研究能够提升多模态学习的可复现性与协作性。我主要参与的项目与实验室项目持续开源，欢迎访问 iLearn Lab 并提出宝贵意见。
Below are representative project pages and repositories from my recent works. Feedback, issues, and pull requests are warmly welcome.
以下是我近期代表性工作的项目主页与代码仓库，欢迎反馈、Issue 与 PR。
📝 Publications
📝 Selected Publications
TEMA: Anchor the Image, Follow the Text for Multi-Modification Composed Image Retrieval [Paper] [Project] [Code]
Zixu Li, Yupeng Hu✉, Zhiheng Fu, Zhiwei Chen, Yongqi Li, Liqiang Nie
Air-Know: Arbiter-Calibrated Knowledge-Internalizing Robust Network for Composed Image Retrieval [Paper] [Project] [Code]
Zhiheng Fu, Yupeng Hu✉, Qianyun Yang, Shiqi Zhang, Zhiwei Chen, Zixu Li†
ConeSep: Cone-based Robust Noise-Unlearning Compositional Network for Composed Image Retrieval [Paper] [Project] [Code]
Zixu Li, Yupeng Hu✉, Zhiwei Chen, Mingyu Zhang, Zhiheng Fu, Liqiang Nie
INTENT: Invariance and Discrimination-aware Noise Mitigation for Robust Composed Image Retrieval [Paper] [Project] [Code]
Zhiwei Chen, Yupeng Hu✉, Zhiheng Fu, Zixu Li†, Jiale Huang, Qinlei Huang, Yinwei Wei
FineCIR: Explicit Parsing of Fine-Grained Modification Semantics for Composed Image Retrieval [PDF] [Project] [Code]
Zixu Li, Zhiheng Fu, Yupeng Hu✉, Zhiwei Chen, Haokun Wen, Liqiang Nie
ENCODER: Entity Mining and Modification Relation Binding for Composed Image Retrieval [PDF] [Project] [Code]
Zixu Li, Zhiwei Chen, Haokun Wen, Zhiheng Fu, Yupeng Hu✉, Weili Guan
📝 Challenge Technical Report
R3: Composed Video Retrieval via Reasoning-Guided Recalling and Re-ranking [Technical Report]
Zixu Li, Yupeng Hu✉, Zhiheng Fu, Zhiwei Chen, Weili Guan, Liqiang Nie
TempRet: Temporal Enhancement and Two-Stage Reranking for CVPR 2026 EPIC-KITCHENS-100 Multi-Instance Retrieval Challenge [Technical Report]
Zixu Li, Yupeng Hu✉, Zhiwei Chen, Zhiheng Fu, Xiaowei Zhu, Weili Guan, Liqiang Nie
EgoAdapt: A Multi-Scene Egocentric Adaptation Method for CVPR 2026 HD-EPIC VQA Challenge [Technical Report]
Zhiwei Chen, Yupeng Hu✉, Zixu Li†, Zhiheng Fu, Guozhi Qiu, Weili Guan, Liqiang Nie
OmniEgo-R2: A Routed Reasoning Framework for the 1st Cross-Domain EgoCross Challenge at CVPR 2026 [Technical Report]
Zixu Li, Zhiwei Chen, Zhiheng Fu, Wenbo Wang, Yupeng Hu✉, Weili Guan, Liqiang Nie
EgoAction: Egocentric Action Composition with Reliability-Aware Temporal Fusion for the EPIC-KITCHENS Action Detection Challenge at CVPR 2026 [Technical Report]
Zhiheng Fu, Zixu Li†, Zhiwei Chen, Fangxu Liu, Yupeng Hu✉, Weili Guan, Liqiang Nie
🏆 Honors and Awards
- 2025.10, National Scholarship (国家奖学金).
- 2025.10, Grand Prize in the CICAS Smart Power Scenario Competition.
🏆 Competition
- 1st place 🏅, CVPR VidLLMs Workshop, Reasoned-Aware Composed Video Retrieval Challenge, 2026.
- 1st place 🏅, CVPR EgoVis Workshop, HD-EPIC Challenge, 2026. [Link]
- 1st place 🏅, CVPR EgoVis Workshop, EPIC-KITCHENS Challenge-Multi-Instance Retrieval Track, 2026. [Link]
- 2nd place 🥈, CVPR EgoVis Workshop, EgoCross Challenge-Source-Limited Track, 2026. [Link]
- 2nd place 🥈, CVPR EgoVis Workshop, EgoCross Challenge-Open-Source Track, 2026. [Link]
- 3rd place 🥉, CVPR EgoVis Workshop, EPIC-KITCHENS Challenge-Action Detection Track, 2026. [Link]
📖 Educations
- 2022.09 - Present, Undergrad in the School of Software
📃 Services
- Serving as a reviewer for AAAI, IJCNN.