Liang-Chieh (Jay) Chen- Home Page

Liang-Chieh (Jay) Chen Research Scientist, Apple AI/ML Email: lcchen at cs dot ucla dot edu [google-scholar]

About Me

I am currently a Research Scientist at Apple AI/ML, building cutting-edge visual generative models.

I am widely recognized for my contributions to the DeepLab series (co-developed with George Papandreou): DeepLabv1, DeepLabv2, DeepLabv3, DeepLabv3+. Since December 2014, our introduced atrous convolution (also known as convolution with holes or dilated convolution) has become a foundational technique for dense prediction tasks.

Influential DeepLab derivatives include Auto-DeepLab, Panoptic-DeepLab (winning method of the Mapillary Vistas Panoptic Segmentation track at ICCV 2019, and top performer on Cityscapes leaderboards), Axial-DeepLab, ViP-DeepLab, MaX-DeepLab, and kMaX-DeepLab.

I am also known for my collaborative work on MobileNetv2 and MobileNetv3, which have become standards for efficient neural network design on mobile devices.

Previously, I was a Senior Principal Scientist at Amazon in 2025, and a Research Scientist and Manager at ByteDance Research/TikTok from 2023 to 2025. Prior to that, I spent seven years as a Research Scientist at Google Research in Los Angeles (2016–2023). I received my Ph.D. in Computer Science from the University of California, Los Angeles, in 2015, advised by Alan L. Yuille. I received my M.S. in Electrical and Computer Engineering from the University of Michigan, Ann Arbor.

News

🏆 DeepLab Wins 2026 Frontiers of Science Award! Our landmark paper “DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs” has been honored with the 2026 Frontiers of Science Award. Presented by the International Congress of Basic Science (ICBS) under the Artificial Intelligence & Pattern Recognition category, this award recognizes transformative scientific breakthroughs from the past decade. Read more on the ICBS Official Platform.

• DeltaTok has been accepted as CVPR 2026 highlight. This innovative tokenizer compresses VFM frame features into a single, compact token, powering DeltaWorld—a generative world model capable of simulating diverse and plausible future scenarios.

• Interested in computer vision, visual generation, or representation learning? We are looking for motivated Ph.D. students and full-time researchers to join our team. Feel free to contact me to discuss open positions.

Activities

• Area Chair for ICCV 2019, CVPR 2020/2023/2024, ECCV 2020/2024, NeurIPS 2022/2024/2025, ICML 2025/2026.

• Senior/Lead Area Chair: ICCV 2025, CVPR 2025/2026, ECCV 2026, NeurIPS 2026.

• Action Editor for TMLR 2025/2026.

• Reviewer for CVPR, ECCV, ICCV, NeurIPS.

  • Outstanding reviewer in CVPR 2021

  • Top reviewer in NeurIPS 2019

  • Outstanding reviewer in CVPR 2018

Selected Recent Publications

Taming Outlier Tokens in Diffusion Transformers Xiaoyu Wu, Yifei Wang, Tsu-Jui Fu, Liang-Chieh Chen, Zhe Gan, Chen Wei Technical Report. [preprint (arxiv: 2605.05206)]

Large Language Models are Universal Reasoners for Visual Generation Sucheng Ren, Chen Chen, Zhenbang Wang, Liangchen Song, Xiangxin Zhu, Alan Yuille, Liang-Chieh Chen, Jiasen Lu Technical Report. [preprint (arxiv: 2605.04040)]

Autoregressive Image Generation with Masked Bit Modeling Qihang Yu, Qihao Liu, Ju He, Xinyang Zhang, Yang Liu, Liang-Chieh Chen*, Xi Chen* (*equal advising) In International Conference on Machine Learning (ICML), Seoul, South Korea, July 2026. [preprint (arxiv: 2602.09024)] [project website] [code]

A Frame is Worth One Token: Efficient Generative World Modeling with Delta Tokens (highlight) Tommie Kerssies, Gabriele Berton, Ju He, Qihang Yu, Wufei Ma, Daan de Geus*, Gijs Dubbelman*, Liang-Chieh Chen* (*equal advising) In Conference on Computer Vision and Pattern Recognition (CVPR), Denver, Colorado, USA, June 2026. [preprint (arxiv: 2604.04913)] [project website] [code]

Frequency-Aware Flow Matching for High-Quality Image Generation Sucheng Ren, Qihang Yu, Ju He, Xiaohui Shen, Alan Yuille, Liang-Chieh Chen In Conference on Computer Vision and Pattern Recognition (CVPR), Denver, Colorado, USA, June 2026. [preprint (arxiv: 2604.15521)] [code]