Mo Li
Ph.D. Candidate at Tsinghua University.
I am Mo Li (李墨), a second-year Ph.D. candidate at Tsinghua University, jointly supervised by Prof. Ting Cao, Dr. Kai Chen, Prof. Yunxin Liu, and Prof. Heung-Yeung Shum. I got my Bachelor’s degree in Data Science and Big Data Technology from Harbin Institute of Technology, Shenzhen. My research interests include Large Language Models, Agent RL, Long Context, and Data Science. Beyond my research, I am curious about understanding the universe.
I enjoy vibe coding and build practical tools in my spare time: ccblog for AI-driven blog publishing to WeChat Official Account, Understand Everything for transforming code repositories into interactive documentation, and Whisper-Input-Next for intelligent voice transcription. I was also a main contributor to OpenCompass and VLMEvalKit, which are one-stop platforms for large model evaluation.
Other Info: INTP. Started university at 16, Ph.D. at 20. Into phonk music, swimming, skiing, photography, and piano. Favorite anime character: L Lawliet. Feel free to reach out via email.
Academic Service: Reviewer for ACL Rolling Review (ARR) 2025, ICLR 2026, ICML 2026 (Silver Reviewer).
News
| Apr 20, 2026 | Kimi K2.6 released! Main contributor to the Skill Creator — skill creation from documents (PDF / spreadsheet / slides) into reusable assets for agents. |
|---|---|
| Jan 27, 2026 | Kimi K2.5 with Agent Swarm released! Main contributor to the RL training that teaches K2.5 to coordinate up to 100 parallel sub-agents. |
| Jan 26, 2026 | Two papers accepted to ICLR 2026: Sculptor on LLM cognitive agency via active context management, and MMSI-Bench on multi-image spatial intelligence evaluation. |
| Nov 06, 2025 | Kimi K2-Thinking released! Contributed the Agent framework that enables K2-Thinking to orchestrate parallel sub-agents for multi-step task execution. |
| Jun 20, 2025 | Started my internship at Moonshot AI, focusing on research in Agent RL, Multi-Agent Systems, and Long Context technologies. |
Selected Publications (Full List)
- ArXiv
Learning to Commit: Next-Commit Prediction via Online Supervised Contrastive ReflectionarXiv Preprint, 2026 - ICLR
Sculptor: Empowering LLMs with Cognitive Agency via Active Context ManagementIn Proceedings of the International Conference on Learning Representations (ICLR), 2026
Awards
| Outstanding Graduate of Harbin Institute of Technology, 2024 |
| National Encouragement Scholarship, 2021, 2022, 2023 |
| Outstanding Student of Harbin Institute of Technology, 2021, 2022, 2023 |
| Grand Prize in the 10th "Teddy Cup" Data Mining Challenge Contest, 2022 |