Jing Shi  

Bio

I am a research scientist at Adobe Research, San Jose, working on multimodal foundation models. My work focuses on making these models improve after pre-training, judge visual and language outputs reliably, and scale through better architectures and learning dynamics.

Before joining Adobe, I received my CS PhD from the University of Rochester, advised by Chenliang Xu, and my B.E. from the University of Electronic Science and Technology of China.

• Post-training and self-improvement. RL, self-play, and feedback loops for stronger vision-language models.
• Evaluation and judgment. Multimodal judges, reward models, and evaluation protocols.
• Model architecture. Efficient architectures, learning dynamics, and inference-time behavior.

Internship advice: Interested PhD students can email me a CV and a brief research plan.

Service

• Area Chair, NeurIPS 2026
• Golden Reviewer, ICML 2026

Publications

Most recent publications on Google Scholar.
^‡ indicates equal contribution.

Vitæ

Full Resume in PDF.

• Adobe Research 2022 - now

  Research Scientist
  Media Intelligence Lab
• University of Rochester 2017 - 2022

  Ph.D. Student
  Department of Computer Science
• Adobe Research Summer 2021

  Research Intern
  Media Intelligence Lab
• Adobe Research Summer 2019

  Research Intern
  Media Intelligence Lab
• Tencent AI Lab Summer 2018

  Research Intern
  Reinforcement Learning Center
• University of Electronic Sciense and Technology of China 2013 - 2017

  B.E. Student
  Department of Automation Engineering