About me

Lifu Tu is a research scientist at Salesforce AI Research. He earned his Ph.D. from TTI-Chicago, a philanthropically endowed academic computer science institute located on the University of Chicago campus. During his Ph.D. studies, he was advised by Kevin Gimpel.

His current research focuses on grounding for LLMs: multilingual and multimodal generation, retrieval-augmented generation (RAG), and methods to enhance factual consistency in large language model outputs. Beyond language, he explores cross-modal innovation—most notably text-to‑video synthesis (as demonstrated in the xGen‑VideoSyn‑1 model) — and structured prediction techniques for richer, more reliable generation systems.

The following are some of his previous projects: multilingual retrieval, LLM benchmarking, long video generation, code generation, etc.