I am Wei Li, a Ph.D. student at the School of Computer Science, Peking University, expected to graduate in July 2027. I am a member of the Institute of Computational Linguistics, advised by Prof. Houfeng Wang. I am currently a research intern at ByteDance Seed.
My research focuses on Coding Agents, especially optimizing agents for long-horizon tasks and learning from difficult problems. More broadly, I am interested in extending the capabilities of Coding Agents, which I believe are one of the most promising near-term paths toward general-purpose AI systems.
Education
2022.08 - 2027.07
Peking University, School of Computer Science, Ph.D. student in Computer Science
2019.09 - 2022.07
Tsinghua University, School of Economics and Management, B.A. in Economics, second bachelor's degree
2019.05 - 2022.07
Tsinghua University, Department of Automation, B.E. in Automation
2017.09 - 2019.05
Tsinghua University, Mechanical Engineering broad category
Internship Experience
2025.11 - Present
ByteDance Seed, Research Intern, Data for software engineering (SWE)
2024.09 - 2025.09
Microsoft Research Asia, Research Intern, Repository-level coding
2021.07 - 2021.09
MoonShot Team, Recurrent AI, Research Intern, zero-shot learning and prompt tuning
Publications
First-author Publications
ICSME 2026, CCF-BSWE-Ext: Scaling and Extending Augmented Data for Mid-Training of Repository-Level Coding Tasks
Wei Li, X. Zhang, S. Wei, Y. Gao, Z. Guo, W. Luo, F. Song, Y. Huang, and H. Wang.
ACL 2025, CCF-AFEA-Bench: A Benchmark for Evaluating Repository-Level Code Generation for Feature Implementation
Wei Li, X. Zhang, Z. Guo, S. Mao, W. Luo, G. Peng, Y. Huang, H. Wang, and S. Li.
NAACL 2025, CCF-BExplanation based In-Context Demonstrations Retrieval for Multilingual Grammatical Error Correction
Wei Li, W. Luo, G. Peng, and H. Wang.
ACL 2024, CCF-ADetection-Correction Structure via General Language Model for Grammatical Error Correction
Wei Li and H. Wang.
ACM MM 2021, CCF-ATransformer-based Feature Reconstruction Network for Robust Multimodal Sentiment Analysis
Z. Yuan* and Wei Li*, H. Xu, and W. Yu.
Co-author Publications
ICML 2026, CCF-AMeasuring and Mitigating Post-hoc Rationalization in Reverse Chain-of-Thought Generation
G. Peng, Z. Chen, W. Luo, Y. Wen, Wei Li, R. Feng, R. Le, C. Yang, Z. An, et al.
ACL 2026, CCF-ATwo Pathways to Truthfulness: On the Intrinsic Encoding of LLM Hallucinations
W. Luo, G. Peng, Wei Li, S. Wei, F. Song, L. Wang, N. Yang, X. Zhang, J. Jin, et al.
NeurIPS 2025, CCF-ATime: A Multi-level Benchmark for Temporal Reasoning of LLMs in Real-world Scenarios
S. Wei, Wei Li, F. Song, W. Luo, T. Zhuang, H. Tan, Z. Guo, and H. Wang.
arXiv:2510.09535Mitigating Overthinking through Reasoning Shaping
F. Song, S. Wei, B. Gao, Y. Wang, W. Luo, Wei Li, L. Yao, W. Xiong, L. Chen, et al.
ACL 2025, CCF-AOdysseus Navigates the Sirens' Song: Dynamic Focus Decoding for Factual and Diverse Open-ended Text Generation
W. Luo, F. Song, Wei Li, G. Peng, S. Wei, and H. Wang.
ACL 2025, CCF-ALearn to Memorize: Scalable Continual Learning in Semiparametric Models with Mixture-of-Neighbors Induction Memory
G. Peng, T. Ge, W. Luo, Wei Li, and H. Wang.
ACL 2025 FindingsEncode Errors: Representational Retrieval of In-Context Demonstrations for Multilingual Grammatical Error Correction
G. Peng, Wei Li, W. Luo, and H. Wang.
arXiv:2406.07070Halludial: A Large-scale Benchmark for Automatic Dialogue-level Hallucination Evaluation
W. Luo, T. Shen, Wei Li, G. Peng, R. Xuan, H. Wang, and X. Yang.
Interests
Technical
Agent optimization for long-horizon tasks.
Languages
Chinese, English, Japanese (JLPT N2), Korean (TOPIK Level 5), linguistics, and experience learning French, German, Portuguese, Italian, Arabic, and Russian.
Photography & Videography
Photography, videography, and visual storytelling.
Sports
Tennis, cycling, and related outdoor activities.
