I am currently a Algorithm Engineer at Tiktok. I am fortunate to collaborate closely with Zejun MA, Wei Li, and Prof. Ziwei Liu. Previously, I obtained my master’s degree from Beijing University of Posts and Telecommunications, supervised by Prof. Jingyu Wang.
My research focuses on Multimodal Understanding and Reasoning. Recently, I am working on build strong Multimodal Agent through end-to-end Reinforcement Learning.
MMSearch-R1 is an e2e RL framework that enables LMMs to perform on-demand, multi-turn search with real-world multimodal search tools. MMSearch-R1-7B model outperforms same-size traditional RAG baselines and cuts search calls by over 30%.
LLaVA-Video series models are Video-LLMs fully trained on our high-quality synthetic dataset, LLaVA-Video-178K, which comprises 178K video captions and 1.15M video QAs. Our models demonstrate strong performance across 10+ video understanding tasks.
MDR proposed a simple and effective metric to measure the preference of different LLMs for demonstrations and leveraged this metric to improve demonstration retrieval frameworks at inference stage.
TikTok-AIIC, ByteDance, 07/2025 ~ Present
Algorithm Engineer
Working on multimodal understanding, reasoning, and agent systems.
NIRC-Lab, Beijing University of Posts and Telecommunications, 09/2022 ~ 06/2025
M.Eng. in Computer Science and Technology
Worked on time series forecasting, prompting for LLMs.