About Me
I am an Applied Scientist at Amazon Prime Video.
I earned my Ph.D. in Computer Science from
Penn State University (2022–2026), advised by Prof. Abhinav Verma,
and my M.S. (2022) and B.S. (2019) from USTB, advised by Prof. Xu-Cheng Yin.
My current research centers on Causal World Modeling, MLLM Agent Post-Training & Self-Evolution, and also Judge.
After nearly a decade in computer vision, I have grown most curious about a different question: how agents can model
open-ended, creative tasks, capturing human taste and, ultimately, modeling how human society itself evolves.
I believe the next frontier is not perceiving the world more accurately, but reasoning about what people find valuable and worth creating.
Publications
Video/Image Generation
What Happens Next? Next Scene Prediction with a Unified Video Model
Xinjie Li, Zhimin Chen, Rui Zhao, Florian Schiffers, Zhenyu Liao, Vimal Bhat
HiGen @ CVPR 2026 Oral, 2026
Co-Speech Gesture Video Generation with Implicit Motion-Audio Entanglement
Xinjie Li, Ziyi Chen, Xinlu Yu, Iek-Heng Chu, Peng Chang, Jing Xiao
CVPR, 2025
Are Conditional Latent Diffusion Models Effective for Image Restoration?
Yunchen Yuan, Junyuan Xiao, Xinjie Li
Arxiv Preprint, 2024
DiffBody: Human Body Restoration by Imagining with Generative Diffusion Prior
Yiming Zhang, Lionel Z. WANG, Sizhuo Ma, Xinjie Li, Jian Ren, Zhihang Zhong, Jian Wang
ICCP, 2025
SD-GAN: Semantic Decomposition for Face Image Synthesis with Discrete Attribute
Kangneng Zhou, Xiaobin Zhu, Daiheng Zhou, Kai Lee, Xinjie Li, Xu-Cheng Yin
ACM MM, 2022
Video Understanding
CI w/o TN: Context Injection without Task Name for Procedure Planning
Xinjie Li
Arxiv Preprint, 2024
Repetitive Action Counting with Motion Feature Learning
Xinjie Li, Huijuan Xu
WACV, 2024
MEID: Mixture-of-Experts with Internal Distillation for Long-Tailed Video Recognition
Xinjie Li, Huijuan Xu
AAAI Oral, 2023
AI4Science
RetroMoE: A Mixture-of-Experts Latent Translation Framework for Single-step Retrosynthesis
Xinjie Li, Abhinav Verma
IJCAI, 2025
Sub-Clustering for Class Distance Recalculation in Long-Tailed Drug Discovery
Yujia Su*, Xinjie Li*, Lionel Z. Wang
Arxiv Preprint, 2025
Modeling Inverse Ellipsometry Problem via Flow Matching with a Large-Scale Dataset
Yiming Ma*, Jianzhi Teng*, Xinjie Li*, Xin Sun, Zhiyong Wang, Yuzhou Song, Lionel Z. Wang, Bin Chen
Arxiv Preprint, 2024
Others
DANet: Dynamic Attention to Spoof Patterns for Face Anti-Spoofing
Chun-Yu Sun, Song-Lu Chen, Xinjie Li, Feng Chen, Xu-Cheng Yin
ICPR, 2022
Semantic Bilinear Pooling for Fine-Grained Recognition
Xinjie Li, Chun Yang, Song-Lu Chen, Chao Zhu, Xu-Cheng Yin
ICPR, 2020
Stereo CenterNet based 3D Object Detection for Autonomous Driving
Yuguang Shi, Yu Guo, Zhenqiang Mi, Xinjie Li
Neurocomputing