Zhaowei (James) Wang

I am currently a research intern with the Seed VLM team at ByteDance, working on multimodal long-context capabilities for Seed-VL models with 128K and 1M-token contexts. This line of work also covers agentic long-context settings, where models are stress-tested on long-horizon, tool-augmented tasks requiring persistent multimodal memory and reasoning.

I am pursuing my PhD in the Department of Computer Science and Engineering (CSE) at HKUST, supervised by Yangqiu Song, as a recipient of the Hong Kong PhD Fellowship Scheme (HKPFS). Previously, I was a visiting scholar at EdinburghNLP, working with Mark Steedman. I received my MPhil degree from HKUST CSE in August 2023.

My current research focuses on multimodal long-context modeling and agentic long-context settings for large vision-language models. I have worked on several related projects, including MMLongBench, MemLens, MMProLong, MMRetHeads, and Memory-T1.

I am always open to new collaborations. Please feel free to reach out via email.