About me
Hi, Iโm Ziyi ๐, a third-year Ph.D. candidate in Computer Science at the Khoury College elementary of Computer Sciences, Northeastern University, where I am advised by Prof. Dakuo Wang at the NEU HAI Lab.
My research generally focuses on LLM agents ๐ค, with recent emphasis on synthesizing data for tool-use training, post-training with SFT/RL to improve agent capabilities, and personalized simulation of human behavior.
Prior to joining Northeastern, I received my bachelorโs degree in Electronic Information Science and Technology and my masterโs degree in Electronic and Information Engineering from Tsinghua University. During my masterโs studies, I was a member of the MSIIP Lab, where I was advised by Prof. Ji Wu.
๐ฐ News
Sep 2026 Two co-authored papers (Firefly and SpanUQ) have been accepted to NeurIPS 2026! ๐ Congrats and many thanks to all collaborators!
Jul 2026 Reached candidacy! ๐โจ
Apr 2026 Two first-authored papers (OPeRA and Trajectory2Task) are accepted to ACL 2026! ๐
Jan 2026 Co-authored paper Shop-R1 is accepted to ICLR 2026! I'll be presenting in Rio de Janeiro โ see you there! ๐ด
Dec 2025 Traveling to San Diego โ๏ธ and presenting our works (OPeRA, Shop-R1, See-Think-Act, Customer-R1) on Multi-Turn Interaction LLM Workshop / the Scaling Environment Agent Workshop at NeurIPS 2025!
๐ผ Internship
Applied Scientist Intern @ Amazon, Palo Alto/Santa Clara Sep 2025 โ July 2026
Training tool-calling LLM agents with synthetic data.
Applied Scientist Intern @ Amazon, Seattle May โ Aug 2025
Personalized simulation of human behaviors via RL-based LLM agent.
Research Intern @ Xiaomi, Beijing Aug โ Nov 2023
Autonomous driving planning with multimodal large models.