CV

Here are my English and Chinese CV.

Education

• Ph.D. Candidate of Computer Software and Theory, University of Chinese Academy of Sciences, 09/2022 - Present
• Bachelor of Data Science and Big Data Technology, University of Science and Technology of China, 2018 - 2022

Work experience

• 07/2024 - 05/2026: Research Intern
  • Huawei Cloud, Pangu Reinforcement Learning Group
  • Safety Alignment: Conducted multi-objective RLHF on the financial capabilities and security of the Pangu 7B large language model based on the OpenRLHF framework. Maintained the model’s general financial capabilities while enhancing its security, achieving improvements in financial compliance abilities.
  • Latent Reasoning: Developed a token-level routing strategy for large and small language models under latent reasoning mode based on the SGLang framework, which reduces the generation length and improves the generation speed.
  • Reasoning RL: Developing a novel RL method for large language models based on model fusion decoding and self-sampling, in order to alleviate the entropy collapse phenomenon of large language models in math and coding reasoning tasks, enhance model exploration, and improve model reasoning ability.
  • Supervisor: Yunkun Xu
• 10/2021 - 03/2022: Research Intern
  • Zhejiang Yingyang Asset Management Co., Ltd.
  • Stock Factors Mining: Developed a deep learning-based low-frequency stock selection model using Pytorch for data processing and factors mining. Tests on historical market transaction data show a compound return of 1.7%.

Publications

Gradient-Adaptive Policy Optimization: Towards Multi-Objective Alignment of Large Language Models

Chengao Li, Hanyu Zhang, Yunkun Xu, Hongyan Xue, Xiang Ao, and Qing He. 2025. Gradient-Adaptive Policy Optimization: Towards Multi-Objective Alignment of Large Language Models. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 11214–11232, Vienna, Austria. Association for Computational Linguistics.

Controlling large language models through concept activation vectors

Zhang, H., Wang, X., Li, C., Ao, X., & He, Q. (2025). Controlling Large Language Models Through Concept Activation Vectors. Proceedings of the AAAI Conference on Artificial Intelligence, 39(24), 25851-25859.

White Book on Safety Evaluation of Generative Large Models (2024)

Teaching

Cognitive Computing

Data Structure

Data Structure and Database

Data Structure

Skills

• Technical: C/C++, SQL, Python, Distributed training & inference for LLM, Pytorch, OpenRLHF, SGLang, Deepspeed
• Awards:
  • USTC Top Talent Program Scholarship
  • USTC Outstanding Student Scholarship
  • Di Ao Scholarship
  • Zeng Xianzi Scholarship
  • 2nd Prize of Anhui Province in Chinese Mathematics Competition
  • UCAS Merit Student
• Patents: Granted Patent. “A Data Flywheel Fine-Tuning Method and Device” (Patent No. CN120234408A, 2025). Lead Inventor.