PeopleofSuperintelligence

Researcher

Xingchao Liu

Also known as 刘星超

Xingchao Liu is a machine-learning researcher whose work centers on generative modeling and multimodal AI. During his Ph.D. research at the University of Texas at Austin, he studied multi-task learning and developed InstaFlow for one-step text-to-image generation. At DeepSeek, he coauthored DeepSeek-V3, the Janus multimodal models and DeepSeek-R1 research on reinforcement learning for reasoning. As of October 2026, he is a researcher in DeepSeek AI's multimodal group and a coauthor of DeepSeek-V4.1-Flash.

Now at
DeepSeek

Timeline

  1. 2021

    Multi-task learning at UT Austin

    During his Ph.D. research at UT Austin's Statistical Learning and AI group, he coauthors the 2021 NeurIPS paper “Conflict-Averse Gradient Descent for Multi-task Learning” with Bo Liu, Xiaojie Jin, Peter Stone and Qiang Liu.

  2. 2023

    InstaFlow for one-step generation

    In 2023, he coauthors InstaFlow with Xiwen Zhang, Jianzhu Ma, Jian Peng and Qiang Liu, using a text-conditioned pipeline to turn Stable Diffusion into a one-step text-to-image model.

  3. 2024

    Janus separates visual pathways

    Janus is an autoregressive framework he coauthors in 2024; it separates visual encoding paths for multimodal understanding and image generation within one transformer.

  4. 2024

    JanusFlow combines two methods

    In 2024, he coauthors JanusFlow, a single model that combines autoregressive language models with rectified flow for image understanding and generation.

  5. 2024

    The DeepSeek-V3 technical report

    He is among the coauthors of DeepSeek-V3's 2024 technical report, which presents a 671-billion-parameter mixture-of-experts model with 37 billion parameters active per token.

  6. 2025

    Scaling Janus-Pro

    In 2025, he coauthors Janus-Pro, which expands training data and model size to improve multimodal understanding and text-to-image instruction-following.

  7. 2025

    Reasoning with DeepSeek-R1

    He coauthors DeepSeek-R1 in 2025, a paper that shows reinforcement learning can encourage reasoning behaviors in language models without human-labeled reasoning trajectories.

  8. As of October 2026

    Research at DeepSeek AI

    As of October 2026, he is a machine-learning researcher in DeepSeek AI's multimodal group. His work spans generative modeling and multimodal intelligence, and he is among the authors of the 2026 DeepSeek-V4.1-Flash report on a 552-billion-parameter multimodal model with contexts up to one million tokens.

Career

  1. NowDeepSeekMachine-Learning Researcher

Education

  1. University of Texas at AustinPhD

From the biography.

  1. The median h-index is 48 at universities and 14 at frontier AI labsIncludes DeepSeek. The median h-index is 48 at universities and 14 at frontier AI labs. We compare publishing careers, recent citations, dual appointments and job moves across 2,420 university researchers and 308 frontier-lab researchers.
  2. Companies supply the most prior-work hires at 7 of the 9 frontier AI labsIncludes DeepSeek. Our estimates of the organizations where 550 people worked directly before joining 9 frontier AI labs across 620 career transitions, comparing the main feeder organizations per lab.
  3. 5,000 frontier users would be 38.2% of the core labs' 13,100 reported staffIncludes DeepSeek. 5,000 frontier users would be 38.2% of the core labs' 13,100 reported staff. We test the estimate against 3,913 named model contributors and dated all-staff records across Google DeepMind, OpenAI and Anthropic, plus six additional frontier-grade labs.
  4. Google alumni have founded or led 133 other organizations, more than any other AI labIncludes DeepSeek. We estimate that alumni of 17 major AI labs went on to found or lead 350 other organizations, including 226 they founded, and we trace 472 career moves directly between 23 labs since 2010.

Sources and credits

Photo: We have no published photo of Xingchao Liu yet. Send us a free-licensed photo.

  1. Xingchao Liu - Personal Website gnobitab.github.io. Checked 20 October 2026.
  2. AI-Lab People - Xingchao Liu cs.utexas.edu. Checked 20 October 2026.
  3. InstaFlow: One Step is Enough for High-Quality Diffusion-Based Text-to-Image Generation arxiv.org. Checked 20 October 2026.
  4. Janus: Decoupling Visual Encoding for Unified Multimodal Understanding and Generation arxiv.org. Checked 20 October 2026.
  5. JanusFlow: Harmonizing Autoregression and Rectified Flow for Unified Multimodal Understanding and Generation arxiv.org. Checked 20 October 2026.
  6. DeepSeek-V3 Technical Report arxiv.org. Checked 20 October 2026.
  7. Janus-Pro: Unified Multimodal Understanding and Generation with Data and Model Scaling arxiv.org. Checked 20 October 2026.
  8. DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning arxiv.org. Checked 20 October 2026.
  9. DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression arxiv.org. Checked 20 October 2026.