I’m an undergraduate student at Nanjing University of Science and Technology. During my undergraduate study, I have ranked first for three years and won multiple scholarships. I have been actively engaged in research and published several CCF conference papers. I’m an incoming a PhD student at the Department of Computer Science, Hong Kong University of Science and Technology in fall 2026, supervised by Prof. Shuai Wang.
My research interest includes Trustworthy AI, Large Language Model and Diffusion Model, and I aim to address critical trustworthiness challenges at the intersection of these fields.
🔥 News
- 2025.07: 🎉🎉 I got the offer from the School of Computer Science, Peking University at the summer camp.
- 2025.06: 🎉🎉 Our paper CtrlDiff is online at Arxiv.
- 2025.06: 🎉🎉 Our paper ScoreAdv is online at Arxiv.
- 2024.12: 🎉🎉 Our paper is accepted by AAAI2025, congratulations to our co-authors!
📝 Publications
- [Arxiv] CtrlDiff: Boosting Large Diffusion Language Models with Dynamic Block Prediction and Controllable Generation, Chihan Huang, Hao Tang.
- [Arxiv] ScoreAdv: Score-based Targeted Generation of Natural Adverarial Examples via Diffusion Models, Chihan Huang, Hao Tang, code.
HUANG: A Robust Diffusion Model-based Targeted Adversarial Attack Against Deep Hashing Retrieval [Code]
Chihan Huang, Xiaobo Shen
Deep hashing models excel at retrieval but remain vulnerable to adversarial attacks. While most attacks target classification, few address retrieval models. We propose HUANG, the first black-box targeted attack for hashing retrieval using a diffusion model. Our method shifts from benign to adversarial distributions via perturbations and a residual image. Experiments show HUANG achieves state-of-the-art performance in black-box targeted attacks, with strong robustness and transferability due to the dynamic interplay between denoising and perturbation addition.
PoemBERT: A Dynamic Masking Content and Ratio Based Semantic Language Model For Chinese Poem Generation [Code]
Chihan Huang, Xiaobo Shen
Ancient Chinese poetry lacks dedicated pre-trained models. We propose PoemBERT, a BERT model for classical poetry. It uses sentiment and pinyin embeddings, plus novel masking strategies, to better capture poetic emotion, imagery, and linguistic nuances. Fine-tuned on tasks like generation and classification, PoemBERT achieves state-of-the-art results in both automated and human evaluations.
Efficient Multi-branch Black-box Semantic-aware Targeted Attack Against Deep Hashing Retrieval [Code]
Chihan Huang, Xiaobo Shen
Deep hashing excels at retrieval but is vulnerable to adversarial attacks, especially in targeted black-box scenarios. We propose EmbSTar, an efficient multi-branch attack framework. It distills a knockoff model and employs novel Target Fusion and Adaptation modules to semantically align adversarial images with a target, enabling powerful, query-efficient attacks. Experiments confirm EmbSTar’s state-of-the-art performance.
🎖 Honors and Awards
- 2025.12 BYD Scholarship (Top 0.1%)
- 2024.09 Top Moral Student of Nanjing University of Science and Technology (Top 0.3%)
- 2024.09 Superior Scholarship of Nanjing University of Science and Technology
- 2024.04 First-Class Scholarship of Nanjing University of Science and Technology
- 2023.09 Excellent Scientific Student of Nanjing University of Science and Technology
- 2023.04 First-Class Scholarship of Nanjing University of Science and Technology
📖 Educations
Hong Kong University of Science and Technology
Coming soon.
- Coming soon.
- Coming soon.
- Coming soon.
Nanjing University of Science and Technology
Won 10 national awards and published 3 CCF papers.
- Awarded BYD Scholarship (Top 0.1%) and Top Moral Student (Top 0.3%).
- Got the offer from School of Computer Science, Peking University.
- Won National First Prize in MCM/ICM, ROBOCON, NECCS, etc.
💬 Invited Talk
🔧 Services
Conference Reviewer
- AAAI 2026
- NeurIPS 2025
- IJCAI 2025
- IJCNN 2025
- ICWSM 2025
Journal Reviewer
- IEEE Transactions on Multimedia
- Neurocomputing
- Applied Intelligence
- Pattern Recognition Letters
💻 Internships
- July. 2025 - Sep. 2025, Research Assistant at Zayed University, UAE.
- Jan. 2025 - Jun. 2025, Research Intern at School of Computer Science, Peking University, China.
😋 Interests
Outside of research, I enjoy playing badminton, singing, watching TV shows and travelling. I rarely play video games. Some of my favorite singers include Jay Chou, JJ Lin, Jason Zhang, Joker Xue, Shen Zhou, G.E.M, Ghost Huang and Feige Zhou. That said, I listen to a wide range of music and mostly Mandopop.
My taste in shows is also quite diverse. I like British dramas like Sherlock, American series like The Good Doctor, and Marvel shows such as Agents of S.H.I.E.L.D., What If…?, and WandaVision. I also watch Chinese dramas like Lighter and Princess, The Knockout, and Blossoms Shanghai. I enjoy music variety shows too, such as Singer, The Treasured Voice, and Melody Journey.
If you play badminton, hit me up! I’m up for men’s singles, men’s doubles, or mixed doubles (barely tried mixed but always open to trying!).
I’m fond of travelling and I’ve visited 18 countries so far:
- Asia (6): Cambodia, China, Malaysia, Thailand, Turkey, Vietnam
- Europe (11): France, Greece, Iceland, Italy, Monaco, Russia, Spain, Sweden, Switzerland, The United Kingdom, Vatican
- America (1): USA