Peking University
Sep 2023 – Jun 2027 (expected)Dual degree in Computer Science
School of Mathematical Sciences
About
I am an undergraduate at Peking University, pursuing a dual degree in Mathematics and Computer Science. I expect to graduate in June 2027.
My research focuses on large language models, including agentic reasoning, reinforcement learning and post-training, safety alignment, and decoding for language diffusion models.
Outside research, I share course notes and small projects on my blog, and enjoy listening to JJ Lin.
News
- 1 paper accepted by EMNLP 2026. Publication details
- I joined Optiver as a quantitative trader intern.
- We open-sourced SearchSwarm, a multi-agent framework for long-horizon deep research. GitHub
- I joined the BaiLing Foundation Model Team at Ant Group as an LLM post-training research intern.
- 1 paper accepted by ICML 2025 Workshop on Models of Human Feedback for AI Alignment. Publication details
Education
Shanghai High School
Sep 2020 – Jun 2023High school
Publications
* Equal contribution.
SearchSwarm: Towards Delegation Intelligence in Agentic LLMs for Long-Horizon Deep Research
Submitted to EMNLP 2026; under review.
A multi-agent framework that uses sub-agents for active context management in long-horizon deep research, with supervised fine-tuning data for learning delegation.
Decoding Large Language Diffusion Models with Foreseeing Movement
EMNLP 2026.
A search-based decoding method that combines local and global information for discrete optimization, with an accelerated variant that concentrates search on selected decoding steps.
Advancing LLM Safe Alignment with Safety Representation Ranking
ICML 2025 Workshop: Models of Human Feedback for AI Alignment.
A safety-alignment method that ranks candidate responses using intermediate Transformer representations and a lightweight similarity scorer.
Awards
- Exemplary Student LeaderPeking University2025
- National ScholarshipPeking University2024
- Outstanding StudentPeking University2024
- Silver Medal, 37th Chinese Mathematical Olympiad (CMO)2021
- First Prize, China Mathematical Olympiad, Shanghai Division2021, 2022
- First Prize, National Olympiad in Informatics in ProvincesShanghai Senior High School Division2019
Service
President, Student Union
Mar 2025 – Mar 2026School of Mathematical Sciences, Peking University
Head of External Relations, Student Union
Mar 2024 – Mar 2025School of Mathematical Sciences, Peking University
Projects
SearchSwarm
2026An open-source multi-agent framework for long-horizon deep research. Trains research agents to delegate subtasks and use compact, evidence-grounded reports for active context management.
LLM agents · Deep research · Supervised fine-tuning
Deep Reinforcement Learning for Chinese Official Mahjong
Dec 2025 – Jan 2026Built an agent trained through behavior cloning and self-play PPO, with an asymmetric actor-critic model and IMPALA-style asynchronous training. Ranked 5th out of 28 teams in the final round of the Botzone competition.
Reinforcement learning · Self-play · Multi-process training
Maintenance and updates to a campus venue-booking tool.
Python · Campus tools