Shihan Dou (窦士涵)

I am a M.Sc student at Fudan NLP Lab at Fudan University, advised by Prof. Tao Gui, Prof. Qi Zhang and Prof. Xuanjing Huang. Before that, I received my B.E. in Information Security from Huazhong University of Science and Technology in 2021.

Currently, I focus on Natural Language Processing, particularly in reward modeling, model alignment (RLHF) and pre-trained language models.

Projects
The RLHF part of Moss (Moss-RLHF) led by Shihan Dou* and Rui Zheng. Moss-RLHF is the second phase of Moss, which helps researchers to align models with humans by RLHF(reinforcement learning from human feedback) more easily and stable. Moss-RLHF includes an open source reward model, a full training framework and a technical report with extensive analysis. We have done extensive experiments to explore the technical route of Reinforcement Learning in Large Language Model for the first time. Models that have been aligned by RLHF can understand human problems better and can be made more helpful and harmless.
The whole of project Moss led by Tianxiang Sun and Xipeng Qiu. Moss is a conversational language model like ChatGPT. It is capable of following users' instructions to perform various natural language tasks including question answering, generating text, summarzing text, generating code, etc. Moss is also able to challenge incorrect premises, and reject inappropriate requests. Here is a brief introduction to Moss.
Project Contributor. TextFlint is a multilingual robustness evaluation platform for natural language processing , which unifies text transformation, sub-population, adversarial attack,and their combinations to provide a comprehensive robustness analysis. So far, TextFlint supports 13 NLP tasks. TextFlint fully covers transformation types, including 20 general transformations, 8 subpopulations and 60 task-specific transformations, as well as thousands of their combinations.
Education & Experience
2023 -
Internship at MiHoYo NLP Group.
2023.02 - 2023.05
Internship at Bytedance TikTok NLP Group.
2021 -
I am a M.Sc student at Fudan NLP Lab at Fudan University, advised by Prof. Tao Gui, Prof. Qi Zhang and Prof. Xuanjing Huang. Up until 2022, my research interests were in robustness and debias in Natural Language Processing. Since the end of 2022, I focus on Large language models training methods, particularly in reward modeling, model alignment (RLHF) and pre-trained language models.
2017 - 2021
B.E. at Huazhong University of Science and Technology with a major in computer science and information security. This is where I first got into the intersection of deep learning and security. In School of Cyberspace Security of HUST, I attended Deqing Zou's group. There I met Dr. Yueming Wu , one of the most important senior (partner) in my undergraduate years. Dr. Yueming Wu taught me a lot about deep learning and security, which helped me get into this field and understand security much faster.
Publications (*: Co-first author)
(MOSS, Technical report)
“Secrets of RLHF in Large Language Models Part I: PPO”
Rui Zheng*, Shihan Dou*, Songyang Gao*, Yuan Hua, Wei Shen, Binghai Wang, Yan Liu, Senjie Jin, Qin Liu, Yuhao Zhou, Limao Xiong, Lu Chen, Zhiheng Xi, Nuo Xu, Wenbin Lai, Minghao Zhu, Cheng Chang, Zhangyue Yin, Rongxiang Weng, Wensen Cheng, Haoran Huang, Tianxiang Sun, Hang Yan, Tao Gui, Qi Zhang, Xipeng Qiu and Xuanjing Huang
(ACL'2023, CCF-A)
“Detecting Adversarial Samples through Sharpness of Loss Landscape,” in Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics.
Rui Zheng*, Shihan Dou*, Yuhao Zhou, Qin Liu, Tao Gui, Qi Zhang, Zhongyu Wei, Xuanjing Huang, Menghan Zhang
(ACL'2023, CCF-A)
“On the Universal Adversarial Perturbations for Efficient Data-free Adversarial Detection,” in Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics.
Songyang Gao*, Shihan Dou*, Qi Zhang12, Xuanjing Huang12, Jin Ma, Ying Shan
(ACL'2023, CCF-A)
“DSRM: Boost Textual Adversarial Training with Distribution Shift Risk Minimization,” in Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics.
Songyang Gao, Shihan Dou, Yan Liu, Xiao Wang, Qi Zhang12, Zhongyu Wei, Jin Ma, Ying Shan
(EMNLP'2022, CCF-B)
“Kernel-Whitening: Overcome Dataset Bias with Isotropic Sentence Embedding,” in Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing
SongYang Gao*, Shihan Dou*, Qi Zhang, Xuanjing Huang
(COLING'2022, CCF-B)
“Decorrelate Irrelevant, Purify Relevant: Overcome Textual Spurious Correlations from a Feature Perspective,” in Proceedings of the International Conference on Computational Linguistics.
Shihan Dou*, Rui Zheng*, Ting Wu, SongYang Gao, Junjie Shan, Qi Zhang, Yueming Wu, and Xuanjing Huang
(ACL'2022, CCF-A)
“MINER: Improving Out-of-Vocabulary Named Entity Recognition from an Information Theoretic Perspective,” in Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics.
Xiao Wang, Shihan Dou, Limao Xiong, Yicheng Zou, Qi Zhang, Tao Gui, Liang Qiao, Zhanzhan Cheng, Xuanjing Huang
(ESEC/FSE'2023, CCF-A)
“Gitor: Scalable Code Clone Detection by Building Global Sample Graph,” in Proceedings of the ACM Joint European Software Engineering Conference and Symposium on the Foundations of Software Engineering.
Junjie Shan*, Shihan Dou*, Yueming Wu, Hairu Wu, and Yang Liu
(TDSC'2022, CCF-A)
“Contrastive Learning for Robust Android Malware Familial Classification,” IEEE Transactions on Dependable and Secure Computing.
Yueming Wu*, Shihan Dou*, Deqing Zou, Wei Yang, Weizhong Qiang, and Hai Jin
(ICSE'2022, CCF-A)
“VulCNN: An Image-inspired Scalable Vulnerability Detection System,” in Proceedings of the IEEE/ACM International Conference on Software Engineering.
Yueming Wu, Deqing Zou, Shihan Dou, Wei Yang, Duo Xu, and Hai Jin
(TOSEM'2021, CCF-A)
“IntDroid: Android Malware Detection Based on API Intimacy Analysis,” ACM Transactions on Software Engineering and Methodology.
Deqing Zou, Yueming Wu, Siru Yang, Anki Chauhan, Wei Yang, Jiangying Zhong, Shihan Dou, and Hai Jin
(ASE'2020, CCF-A)
“SCDetector: Software Functional Clone Detection Based on Semantic Tokens Analysis,” in Proceedings of the IEEE/ACM International Conference on Automated Software Engineering.
Yueming Wu, Deqing Zou, Shihan Dou, Siru Yang, Wei Yang, Feng Cheng, Hong Liang, and Hai Jin
Awards
2021 - 2022, National Scholarship, Fudan University | 复旦大学国家奖学金
2021, Outstanding Graduate, Huazhong University of Science and Technology | 华中科技大学优秀毕业生
2019 - 2020, National Scholarship, Huazhong University of Science and Technology | 华中科技大学国家奖学金
沪ICP备2021034091号

Powered by Shihan Dou (Ablustrund), thanks to Miyehn (小雨) for providing artistic support