陈沁宇

陈沁宇 Qinyu Chen

Master Student@PKU

Peking University

About Me

Qinyu Chen is a last-year master student from School of Computer Science, Peking University, advised by Prof. Sujian Li. He also received his Bachelor’s degree at Peking University, majoring in computer science, advised by Prof. Zhihong Deng. Qinyu is a member of TANGENT and AIIC. Currently, he is working at DeepSeek.

His research interests include natural language processing, large language models, theories and applications of machine learning and recommender systems.

Contact him at chenqinyu@pku.edu.cn

Interests
  • Natural Language Processing
  • Artificial Intelligence
  • Applied Machine Learning
  • Recommender System
Education
  • MS in Artificial Intelligence, 2025 (estimated)

    Peking University, School of Compute Science

  • BSc in Computer Science, 2022

    Peking University, EECS

Experience

 
 
 
 
 
DeepSeek
Deep Learning Engineer / AGI Research
DeepSeek
March 2024 – Present Beijing

My work includes:

  • Building frontier AI application frameworks
  • Researching large language models’ abilities acquisition from a data-driven perspective
  • Brewing ☕️

It’s quantitative change that leads to qualitative change.

 
 
 
 
 
Microsoft
Applied Scientist Intern @ Bing Ads
Microsoft
August 2023 – February 2024 Beijing

Responsibilities include:

  • Combining LLMs with retrieval system
  • Designing and training new models that improve performance by 30+%
 
 
 
 
 
ByteDance
Algorithm Engineer Intern @ Douyin Search
ByteDance
November 2021 – August 2022 Beijing

Responsible for a series of bi-monthly projects:

  • introduction of multi-modal information to optimize search results
  • probabilistic modeling of user click behavior
  • introducing new user behavioral objectives

Improved CTR by 0.15% in online test.

Publications

Obviously, not very prolific.

(2024). Retrieval-based Full-length Wikipedia Generation for Emergent Events. In ACL2024 (under review).

PDF Cite

(2024). Selecting Large Language Model to Fine-tune via Rectified Scaling Law. In ICML2024 & ICLR2024 Workshop.

PDF Cite

(2023). Exploring In-Context Learning for Knowledge Grounded Dialog Generation. In EMNLP2023.

PDF Cite

(2021). Domain Adaptation via Maximizing Surrogate Mutual Information. In IJCAI2022.

PDF Cite

Contact