PeopleofSuperintelligence

Researcher

Yutong Lin

Also known as 林宇桐, impiga

Yutong Lin is a computer vision researcher whose work ranges from vision Transformers and object detection to video generation and high-definition map construction. They coauthored Swin Transformer, which won the ICCV 2021 Marr Prize, and contributed to Swin Transformer V2 and SimMIM in November 2021. They were first author of a 2022 study on frozen giant vision models and a 2023 DETR paper built around single-scale features and global cross-attention. In 2025, they coauthored Sand.ai’s MAGI-1 video model and the DAMap mapping method at ICCV. As of October 2026, they are a PhD candidate at Xi'an Jiaotong University.

Works
63
Citations
41,533
h-index
24
Badges

Publication numbers from OpenAlex, as of 22 October 2026.

Timeline

  1. March 2021

    Swin Transformer

    In March 2021, they coauthor Swin Transformer with Ze Liu, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin and Baining Guo. The hierarchical vision Transformer computes self-attention in shifted windows and reports 87.3 top-1 accuracy on ImageNet-1K and 58.7 box AP on COCO test-dev.

  2. 2021

    An ICCV Marr Prize

    Swin Transformer wins the ICCV 2021 Marr Prize for best paper. Yutong Lin is the Xi'an Jiaotong University author alongside Ze Liu from the University of Science and Technology of China and Yue Cao and Han Hu from Microsoft.

  3. November 2021

    Swin V2 and SimMIM

    In November 2021, they coauthor Swin Transformer V2, which scales to 3 billion parameters and trains with images up to 1,536×1,536 pixels. They also coauthor SimMIM, a masked image modeling framework that reaches 83.8% top-1 accuracy on ImageNet-1K with ViT-B.

  4. 2022

    Frozen giant vision models

    In 2022, they are first author of “Could Giant Pretrained Image Models Extract Universal Representations?” with Ze Liu, Zheng Zhang, Han Hu, Nanning Zheng, Stephen Lin and Yue Cao. A frozen 3-billion-parameter SwinV2-G backbone reaches 60.0 box mAP on COCO test-dev and 81.7 top-1 accuracy on Kinetics-400.

  5. 2023

    A plain DETR detector

    In 2023, they are first author of “DETR Does Not Need Multi-Scale or Locality Design” at ICCV, with Yuhui Yuan, Zheng Zhang, Chen Li, Nanning Zheng and Han Hu. The detector uses a single-scale feature map and global cross-attention, and reaches 63.9 mAP with a Swin-L backbone after Object365 pretraining.

  6. May 2025

    MAGI-1 video generation

    In May 2025, they coauthor Sand.ai’s MAGI-1, a world model that generates video by autoregressively predicting consecutive frame chunks. Its largest variant has 24 billion parameters and supports context lengths up to 4 million tokens; they are among its core contributors.

  7. 2025

    High-definition map construction

    In 2025, they coauthor DAMap at ICCV, a high-definition map construction method that uses distance-aware focal loss, a hybrid loss scheme and task-modulated deformable attention. They are affiliated with Xi'an Jiaotong University's Institute of Artificial Intelligence and Robotics.

  8. As of October 2026

    PhD candidate at Xi'an Jiaotong

    As of October 2026, they are a PhD candidate at Xi'an Jiaotong University in Xi'an. Their GitHub profile, impiga, pins MAGI-1, Swin Transformer object-detection code and Plain-DETR.

Career

  1. 2025–2026Fuzhou University
  2. 2020–2022Microsoft Research Asia
  3. 2019–2022Harbin Normal University

Education

  1. Xi'an Jiaotong UniversityPhD Candidate

From the biography and OpenAlex affiliations (the years of papers that list the organization), so some years are approximate.

  1. 16% of AI career moves in the 2020s went from a university to a companyIncludes Yutong Lin. We estimate that 16% of AI career moves in the 2020s went from a university to a company; 29% of moves starting at a university went to a company, and 76% ended the university role.
  2. The Massachusetts Institute of Technology trained or employed 44 of the 187 AI major-prize winners, more than any other universityIncludes Yutong Lin. The Massachusetts Institute of Technology trained or employed 44 of the 187 AI major-prize winners, more than any other university. We compare doctorate and 5+ year employment links, career timing and current workplaces for 187 major-prize and 1,111 best-paper winners.
  3. Among employers with 30+ full-time cases, OpenAI leads primary technical role H-1B pay at $310,000, and Amazon, Google and Microsoft file 57% of all casesIncludes Microsoft. Across 440,669 certified US H-1B filings from 44 employers (FY2020–FY2026), OpenAI leads primary technical role base pay at $310,000 and xAI leads overall base pay at $290,000.
  4. Google had the most people shaping AI every year from 2014 to 2025Includes Microsoft. Our estimates of which organizations employed the most people shaping AI each year from 1955 to 2025, tracing the shift from university labs to companies.

Colleagues

People named in this biography, and people with dated positions at the same organization in the same years.

Sources and credits

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

  1. Swin Transformer: Hierarchical Vision Transformer using Shifted Windows arxiv.org. Checked 22 October 2026.
  2. Microsoft Asia’s Swin Transformer V2 Scales the Award-Winning ViT to 3 Billion Parameters and Achieves SOTA Performance on Vision Benchmarks syncedreview.com. Checked 22 October 2026.
  3. Swin Transformer V2: Scaling Up Capacity and Resolution arxiv.org. Checked 22 October 2026.
  4. SimMIM: A Simple Framework for Masked Image Modeling arxiv.org. Checked 22 October 2026.
  5. Could Giant Pretrained Image Models Extract Universal Representations? arxiv.org. Checked 22 October 2026.
  6. DETR Doesn't Need Multi-Scale or Locality Design arxiv.org. Checked 22 October 2026.
  7. Yuhui Yuan's publication bibliography csauthors.net. Checked 22 October 2026.
  8. MAGI-1: Autoregressive Video Generation at Scale arxiv.org. Checked 22 October 2026.
  9. Magi-1: Autoregressive Video Generation at Scale (HTML version) arxiv.org. Checked 22 October 2026.
  10. DAMap: Distance-aware MapNet for High Quality HD Map Construction openaccess.thecvf.com. Checked 22 October 2026.
  11. DAMap: Distance-aware MapNet for High Quality HD Map Construction (arXiv HTML) arxiv.org. Checked 22 October 2026.
  12. impiga (Yutong Lin) · GitHub github.com. Checked 22 October 2026.