Gu Yuxian, winner of Tsinghua Special Scholarship, joins DeepSeek, focusing on large model compression and efficiency optimization.

According to Beating monitoring, Gu Yuxian, a Ph.D. graduate from the Department of Computer Science at Tsinghua University and the recipient of the 2025 Graduate Special Scholarship, has officially joined DeepSeek, and his name has appeared in the author list of the DeepSeek V4 paper.

Gu Yuxian’s research primarily focuses on efficiency optimization for large models in the pre-training, model compression, and inference stages, with nearly 5,000 total citations on Google Scholar. His previous representative works include the large-model knowledge distillation method MiniLLM (adopted by platforms such as Google, Alibaba, and NVIDIA), as well as the hybrid-architecture model Jet-Nemotron. When processing 256K ultra-long contexts on H100 GPUs, Jet-Nemotron achieves a generation throughput that is 53.6 times faster than traditional full-attention models, and it outperforms mixture-of-experts models with larger parameter sizes in multiple benchmark tests.