PhD Student at SJTU contributing to supervised fine-tuning and model evaluation for reasoning, search, tool use, and safety alignment at ByteDance.
How media typically covers Wei Li
Wei Li as author
ByteDance's Robix, a unified vision-language model, outperforms GPT-4o and Gemini 2.5 Pro in interactive robot task execution through chain-of-thought reasoning and hierarchical control architecture.
“Co-author of the Robix research paper”