cv

Undergraduate

Institute for Interdisciplinary Information Sciences (IIIS), Tsinghua University

Computer Science

High School

Hangzhou Xuejun High School

Physics Olympiad

Middle School

Hangzhou Wenlan Middle School

LLM-Assisted Dataset Generation with Consistency Evaluation for RLHF

Generate a consistency-weighted dataset using LLMs to augment limited human-labeled preferences for improved reward model training in offline RLHF tasks.

• Instructor: Prof. Zhixuan Fang in IIIS
• Cooperator: Jiaxuan Jiang

Stabilizing Offline Reinforcement Learning with Imitation-Based Assistant Rewards

Introduce an imitation-based assistant reward to improve early-stage training stability in offline reinforcement learning on the D4RL benchmark

• Instructor: Prof. Simon Du in UW
• Cooperator: Xinqi Wang

Reward Optimization of Imitation Learning in Offline Multi-Agent Games from Human Feedback

Integrate human feedback into imitation learning for offline multi-agent tasks, using a reward model to improve performance by filtering and optimizing offline datasets

• Instructor: Prof. Zhixuan Fang in IIIS
• Paper: PDF

Multi-Agent Reinforcement Learning from Human Feedback

Initiated the work of Multi-Agent Reinforcement Learning from Human Feedback (MARLHF), combining theoretical analysis and empirical experiments on preference-only offline datasets and proposing new algorithmic techniques

• Instructor: Prof. Simon Du, Prof. Sham Kakade
• Cooperator: Xinqi Wang, Qiwen Cui, Runlong Zhou
• Repo: Jax-MARLHF
• Paper: Arxiv
• ICLR 2025 Under Review

Research on Communication in Multi-Agent Reinforcement Learning

Introduced truthfulness for agents and large language model for communication in decentralized, partially observable multi-agent reinforcement learning, aimed at improving performance and transferability

• Instructor: Prof. Zhixuan Fang in IIIS
• Cooperator: Jiaxuan Jiang and Kaidi Fu

Research on a Self-Developed Information Eliciting Model

Researched and developed a novel Information Eliciting Model incorporating a thinking hierarchy framework to uncover the truth within diverse perspectives, introducing metrics for measuring response hierarchy

• Instructor: Prof. Zhixuan Fang in IIIS
• Cooperator: Weiliang Wang
• Paper: PDF

New Star Academic Program on Real-Time Translucent Neural Rendering

Developed real-time neural rendering techniques for translucent objects by utilizing deep learning to model subsurface scattering, with superior results compared to existing approximation methods

• Instructor: Prof. Kun Xu in Department of Computer Science and Technology
• Cooperator: Di An and Shuzhen Li

Comprehensive Scholarship in Tsinghua University

Comprehensive Scholarship in Tsinghua University

Scholarship of Institute for Interdisciplinary Information Core Technology, First Prize

Freshman Scholarship, Second Prize

Address at IIIS Opening Ceremony

National Physics Olympiad, Second Prize

Tsinghua Physics Camp, First Prize

National English Proficiency Competition, Second Prize

National Oral English Innovative Proficiency Contest, Second Prize

Implementation and Improvement on a 3D Hierarchical Torus Network Topology and its Corresponding Deadlock-Free Routing Algorithm

• Cooperator: Ran Tao
• Instructor: Prof. Kaisheng Ma

Distributed Computation on Matrix Multiplication with Concurrency

• Cooperator: Xingchen Miu
• Instructor: Prof. Wei Xu and Prof. Mingyu Gao

Implementation of RISC-V Processor Model with High-Performance Cache

• Instructor: Prof. Mingyu Gao

Implementation of a Group of Machine Learning Models Including Random Forest and XGBoost

• Instructor: Prof. Yang Yuan

Implementation of Stochastic Algorithms for 2D Ising Model

• Instructor: Prof. Yukai Wu

Implementation of a Distributed Database Model

• Instructor: Prof. Huanchen Zhang

Multimedia Processing Based on FFmpeg

• Instructor: Prof. Huanchen Zhang