Li Ding | 丁立
Scholar / Github / Twitter / LinkedIn / CV
liding256@gmail.com liding@{umass.edu, mit.edu}
I'm currently at Google ( Mountain View), advancing Gemini's capabilities in multimodal LLMs and generative AI agents.
I received my Ph.D. from UMass Amherst CICS in 2024, under the supervision of Lee Spector. My research focused on AI alignment and open-endedness.
During my Ph.D., I worked closely with Scott Niekum (UMass, UT Austin), Joel Lehman (formerly OpenAI), Jeff Clune (UBC, DeepMind), and Masrour Zoghi (Google Research). Additionally, I gained research experience through internships at Google Research and Meta.
Before Ph.D., I was a full-time researcher at MIT with Lex Fridman and Bryan Reimer, and concurrently a graduate student at MIT CSAIL. I did my master's at Univ. of Rochester with Chenliang Xu.
Research
The goal of my research is to align AI systems with human values and enable them to safely learn in open-ended environments. I develop algorithms to enhance the safety and creativity of generative models through reinforcement learning from human feedback (RLHF), alongside methods to improve the generalization of foundation models.
My work spans multiple domains, including multimodal LLMs, image generators, and robotics. I'm also interested in (and have published in) disciplines such as quantum ML, symbolic regression, and human-computer interaction. Before Ph.D., I worked on deep learning for autonomous driving, cognitive modeling, and video recognition.
Selected Publications
For a complete and up-to-date list of publications, please see Google Scholar.
Pareto-Optimal Learning from Preferences with Hidden Context
Ryan Boldi, Li Ding, Lee Spector, Scott
Niekum
arXiv preprint 2024
[arXiv]
POPL learns Pareto-optimal policies or reward functions in RLHF, addressing hidden contexts such as diverse group preferences without needing group labels, offering safe and fair alignment of agents and LLMs.
Quality Diversity through Human Feedback: Towards Open-Ended Diversity-Driven Optimization
Li Ding, Jenny Zhang ,
Jeff Clune , Lee Spector ,
Joel Lehman
ICML 2024
NeurIPS 2023: ALOE Workshop (Spotlight)
[project page] [arXiv] [demo] [talk] [code] [tutorial]
QDHF learns diversity metrics from human feedback and uses those to optimize for the discovery of novel solutions, enhancing the task-solving capabilities of RL agents and the creativity of generative models.
Value of Temporal Dynamics Information in Driving Scene Segmentation
Li Ding, Jack Terwilliger, Rini Sherony, Bryan Reimer, Lex Fridman
IEEE Transactions on Intelligent Vehicles, 2021
[paper] [arXiv] [MIT DriveSeg
Dataset]
Press coverage: [MIT News] [Forbes] [InfoQ] [TechCrunch]
Misc.
Teaching:
TA for MIT 6.S094: Deep Learning for Self-Driving Cars
(Winter 2018-19).
TA for MIT 6.S099: Artificial
General Intelligence (Winter 2019).
Reviewer:
ICLR, NeurIPS, JMLR, CVPR, ICCV, ECCV, etc.
Open source projects:
google-research/ev3: Meta-learning optimization in JAX.
facebookresearch/d2go: Efficient model
training and deployment on mobile platforms.
pyribs: An open-source library for quality
diversity optimization.
mit-deep-learning: Tutorials and
coding assignments for MIT Deep Learning courses (9k+ stars).
MIT AI Podcast
Helped prepare interview
questions, search for guest speakers, etc. for a podcast hosted by Lex
Fridman about technology, science, and the human condition.
(Ranked #1 on Apple
Podcasts in the technology category, 1M views on YouTube.)
(My personal favorite episode is Tomaso Poggio, highly recommended!)
MIT Robocar Workshop
Instructor for a summer/winter workshop at MIT with Tom Bertalan to college and high school students on building and programming
autonomous robocars.
Last updated: 10/2024
Powered by Skeleton
Design inspired by Jon Barron