π About Me
I am Zeyu Qin (秦泽ι°), a Ph.D. student of Computer Science & Engineering at Hong Kong University of Science and Technology (HKUST). My advisors are Prof. Minhao Cheng and Prof. Nevin L. Zhang. Recently, I also collaborate with Prof. Li Shen, Prof. Ruoyu Sun, and Dr. Yongqiang Chen. My research interest is broadly on AI Alignment, and Safety.
Currently, I am focusing on:
1) How to obtain better and more safe Supervision for oversighting LLMs;
2) Eliciting Honest Latent Knowledge (ELK) of LLMs;
3) Faithful Evaluation of LLMs.
I am very fortunate to have worked with lots of distinguished researchers: Prof. Baoyuan Wu, Dr. Yanbo Fan, Prof. Hongyuan Zha, Dr. Jiancong Xiao, and Prof. Piji Li. I had an amazing four years at LGU with my good friends from Room 310 and Room 224. Miss you, guys!
π News
- 2024.09: One Spotlight has been accepted by NeurIPS 2024: Understanding Superficial Robustness of Safety Tuningβthe Backdoor Purification Case.
- 2024.07: I will go to Vienna to attend ICML 2024.
- 2023.10: I will go to New Orleans to attend NeurIPS 2023.
- 2023.09: Two papers (one Spotlight) have been accepted by NeurIPS 2023: 1. Feature shift tuning which achieves SOTA purification performance against backdoor attacks; 2. Spotlight! Imitation learning from imperfect demonstrations (See Ziniu Li).
- 2023.05: New work has been accepted by KDD 2023: The first study about robustness from personalization in FL against backdoor attacks.
π Publications
* denoting equal contribution ^ denoting corresponding author
-
Uncovering, Explaining, and Mitigating the Superficial Safety of Backdoor Defense (Spotlight!)
Rui Min*, Zeyu Qin* ^, Nevin Zhang, Li Shen, Minhao Cheng.
NeurIPS 2024 [arxiv] [code] -
Towards Stable Backdoor Purification through Feature Shift Tuning
Rui Min*, Zeyu Qin* ^, Li Shen, Minhao Cheng.
NeurIPS 2023 [arxiv] [OpenReview] [code] -
Imitation Learning from Imperfection: Theoretical Justifications and Algorithms (Spotlight!)
Ziniu Li*, Tian Xu*, Zeyu Qin, Yang Yu, Zhiquan Luo.
NeurIPS 2023 [OpenReview]
(This is excellent work from Ziniu and Tian. I only conducted part of the experiments. I almost have no idea about Imitation Learning π.) -
Revisiting Personalized Federated Learning: Robustness Against Backdoor Attacks
Zeyu Qin, Liuyi Yao, Daoyuan Chen, Yaliang Li, Boling Ding, Minhao Cheng.
KDD 2023 [arxiv] [code] -
Beyond Factuality: A Comprehensive Evaluation of Large Language Models as Knowledge Generators
Liang Chen, Yang Deng, Yatao Bian, Zeyu Qin, Bingzhe Wu, Tat-Seng Chua, Kam-Fai Wong.
EMNLP 2023 [arxiv] [code] -
Adaptive Smoothness-weighted Adversarial Training for Multiple Perturbations with Its Stability Analysis.
Jiancong Xiao, Zeyu Qin, Yanbo Fan, Baoyuan Wu, Jue Wang, Zhi-Quan Luo.
ICML 2023 Workshop AdvML-Frontiers, 2023 [arxiv] -
Boosting the Transferability of Adversarial Attacks with Reverse Adversarial Perturbation
Zeyu Qin*, Yanbo Fan*, Yi Liu, Li Shen, Yong Zhang, Jue Wang, Baoyuan Wu.
NeurIPS 2022 [arxiv] [OpenReview] [code] -
Random Noise Defense Against Query-Based Black-Box Attacks
Zeyu Qin, Yanbo Fan, Hongyuan Zha, Baoyuan Wu.
NeurIPS 2021 [arxiv] [OpenReview] [code]
π Preprints
* denoting equal contribution
-
Entropic Distribution Matching in Supervised Fine-tuning of LLMs: Less Overfitting and Better Diversity
Ziniu Li, Congliang Chen, Tian Xu, Zeyu Qin, Jiancong Xiao, Ruoyu Sun, Zhi-Quan Luo.
Arxiv, 2024. [arxiv] -
Step-On-Feet Tuning: Scaling Self-Alignment of LLMs via Bootstrapping
Haoyu Wang, Guozheng Ma, Ziqiao Meng, Zeyu Qin, Li Shen, Zhong Zhang, Bingzhe Wu, Liu Liu, Yatao Bian, Tingyang Xu, Xueqian Wang, Peilin Zhao.
Arxiv, 2024. [arxiv]
π Honors and Awards
- 2024.09, NeurIPS 2024 Spotlight Paper
- 2023.09, NeurIPS 2023 Spotlight Paper
- 2021.04, Poster Runner-Up Reward of The First Doctoral and Postdoctoral Academic Forum of Shenzhen Research Institute of Big Data
π Educations
- 2022.08 - Now, Ph.D. student in Computer Science & Engineering, The Hong Kong University of Science and Technology.
- 2018.08 - 2022.05 (Ph.D. β> M.Phil), M.Phil in Computer and Information Engineering, The Chinese University of Hong Kong, Shenzhen.
- 2014.09 - 2018.06, B.Eng. in Information Engineering, Nanjing University of Aeronautics and Astronautics.
π¬ Invited Talks
- 2022.10, Adversarial Transferability in AI Times Forum.
π» Internships
- Summer 2022: Research Intern, Alibaba Damo Academy, Hangzhou, China
- July 2021 β May 2022: Research Intern, Tencent AI Lab, Shenzhen, China
- January 2020 β October 2020: Research Intern, Tencent AI Lab, Shenzhen, China
π§ Contact
zeyu.qin@connect.ust.hk ; zeyu6181136@gmail.com