Anjian Li
Hi! I am a 5th year PhD candidate in the Department of Electrical and Computer Engineering at Princeton University, advised by Prof. Ryne Beeson. I also work closely with Prof. Adji Bousso Dieng and Prof. Bartolomeo Stellato. My research focuses on generative AI and world models for safe, reliable reasoning and decision-making in complex environments, with applications to autonomous systems and robotics. I’m also interested in developing machine learning methods to accelerate non-convex optimization.
Over the last summer, I interned at Waymo and worked on post-training a generative foundation model running on thousands of fully autonomous vehicles.
Previously I did an internship at Honda Research Institute USA, working on consistency models. I did my master in computing science at Simon Fraser University, advised by Prof. Mo Chen. Before that, I received my bachelor in mathematcis at Beijing Normal University.
news
| Jun 18, 2026 | Our paper Recurrent Autoregressive Diffusion: Global Memory Meets Local Attention, which brings compact recurrent memory to long-video generation world models, has been accepted to the European Conference on Computer Vision (ECCV 2026)! |
|---|---|
| Jan 15, 2026 | Our paper Adaptive Time Step Flow Matching for Autonomous Driving Motion Planning, which adapts inference steps online for real-time autonomous driving motion planning, has been accepted to the 2026 IEEE Intelligent Vehicles Symposium (IV 2026)! |
| Jul 01, 2025 | Our paper Predictive Planner for Autonomous Driving with Consistency Models, which enables fast joint prediction and planning for interactive driving, has been accepted to the 2025 IEEE 28th International Conference on Intelligent Transportation Systems (ITSC 2025)! |
| Jun 02, 2025 | I started a summer internship at Waymo, working on generative foundation models for planning! |
| Feb 28, 2025 | Our paper DiffuSolve: Diffusion-Based Solver for Non-Convex Trajectory Optimization, which uses diffusion models to warm-start challenging trajectory optimization problems, has been accepted to the 7th Annual Learning for Dynamics & Control Conference (L4DC)! |
selected publications
-
Recurrent Autoregressive Diffusion: Global Memory Meets Local AttentionIn European Conference on Computer Vision (ECCV), 2026 -
Predicting Long-Term Human Behaviors in Discrete Representations via Physics-Guided DiffusionIn IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2024 -
Guaranteed-Safe Approximate Reachability via State Dependency-Based DecompositionIn American Control Conference (ACC), 2020
