CV
Last updated: July 2026
Education
M.Eng. in Electronic Information (Computer Science and Technology), Tsinghua University, Expected June 2028
- Advisor: Associate Professor Jianfei Chen
- Research group: TSAIL
B.Eng. in Computer Science and Technology, Tsinghua University, June 2026
- Relevant coursework (grade A or above): Ordinary Differential Equations; Probability and Statistics; Fundamentals of Programming; Programming and Training; Software Engineering; Operating Systems; Principles of Signal Processing; Database Special Topic Training
- Previous major: Medicine. Transitioned to Computer Science to pursue a passion for artificial intelligence.
Research Statement
- My research focuses on post-training and test-time scaling for large language models (LLMs), with the goal of improving model and agent capabilities in complex settings such as long-context reasoning, long-form generation, coding, and multi-agent collaboration.
Publications
- Scaling External Knowledge Input Beyond Context Windows of LLMs via Multi-Agent Collaboration
Zijun Liu*, Zhennan Wan*, Peng Li, Ming Yan, Fei Huang, Yang Liu
ACL 2026 Main Conference, co-first author. [Paper] [Code] [Data]
Research Experience
- September 2025 - Present: Research Member, Tsinghua Statistical Artificial Intelligence and Learning Group (TSAIL), Tsinghua University
- LC-Zero: Self-Evolving Long-Context Reasoning from Zero Data
- Developing a zero-data synthesis pipeline that constructs short “worlds” containing questions, answers, and evidence, then expands them into long contexts.
- The model learns from self-synthesized text to preserve correctness while continually increasing task difficulty; GRPO is used to enable self-evolving training (ongoing).
- Scaling Long-Form Story Generation with Structured Narrative State Tracking
- Proposed a tool-using agent framework that maintains structured narrative states to improve consistency in long-form fiction.
- Scaled story generation from 10,000 to 100,000 words without noticeable performance degradation.
- LC-Zero: Self-Evolving Long-Context Reasoning from Zero Data
- September 2024 - May 2025: Research Intern, Natural Language Processing and Social Humanities Computing Lab (THUNLP), Tsinghua University
- Scaling External Knowledge Input Beyond Context Windows of LLMs via Multi-Agent Collaboration
- Developed ExtAgents, a multi-agent framework that distributes external knowledge across agents and enables intensive inter-agent communication, overcoming the input limits of a single LLM context window.
- Matched or outperformed conventional approaches on knowledge-intensive question answering while achieving greater parallelism and scalability.
- Scaling External Knowledge Input Beyond Context Windows of LLMs via Multi-Agent Collaboration
- February 2024 - August 2024: Research Intern, 3D Visual Computing and Machine Intelligence (3DVICI) Lab, Institute for Interdisciplinary Information Sciences, Tsinghua University
- Editing Human Videos for Robotic Skill Training
- Used video editing techniques to convert human hand-object interaction videos into robotic hand-object interaction videos.
- Extracted 6D poses from human videos and compared human and robot data, referencing pipelines such as OpenVLA to evaluate the usefulness of human videos for robot training.
- Editing Human Videos for Robotic Skill Training
Internship Experience
- February 2026 - May 2026: LLM Algorithm Intern, TRAE Intelligent Coding Algorithms, ByteDance
- Seed Code Model - Coding Capability Optimization
- Systematically evaluated long-context coding performance on LoCoBench, LoCoDiff, and other benchmarks to identify gaps versus leading models.
- Analyzed failures in code-state tracking and synthesized training data from online trajectories, improving benchmark performance by approximately 8% to match GLM-5.
- Designed a rubric-based evaluation framework for open-ended technical QA and conducted targeted synthetic-data supervised fine-tuning, narrowing the gap to GPT-5.2 by approximately 49%.
- Seed Code Model - Coding Capability Optimization
- June 2025 - August 2025: Applied Research Intern, Code Intelligence Center, Technology and Engineering Group (TEG), Tencent
- CodeLLM Applications for Cursor Prediction and Smart Rewrite - Project Lead
- Synthesized data for 5 cursor-prediction and 15 smart-rewrite scenarios, producing 21K high-quality training examples.
- Conducted supervised fine-tuning and evaluation across Qwen2.5-Coder 0.5B/3B/7B/14B models; iterative improvements outperformed GPT-4.1 on selected tasks.
- Built a VS Code extension with real-time inference and visual interaction; the project ranked 1st among 5 teams.
- CodeLLM Applications for Cursor Prediction and Smart Rewrite - Project Lead
Course Project
- April 2025 - June 2025: Frontiers in AI Safety and Governance (Spring 2025)
- Adaptive Safety Priming: Inference-Time Safeguards for Large Reasoning Models
- Developed Adaptive Safety Priming (ASP), a lightweight, dynamic safety mechanism that leverages the step-by-step inference process of large reasoning models to enable real-time intervention. This approach provides a more adaptive and resource-efficient path toward robustly safe models. [Report]
- Adaptive Safety Priming: Inference-Time Safeguards for Large Reasoning Models
Skills
- Programming: C, C++, Python
- Machine learning: PyTorch, vLLM, VeRL
- Tools: Git, Linux, Docker, LaTeX
- English: College English Test Band 6 (CET-6): 601/710
Awards
- Academic Excellence Scholarship, 2024-2025 Academic Year
- Tsinghua University Software Engineering Outstanding Project Award, 2024