Senior Research Scientist — World Models & Closed-loop Simulation
Yuna Kang
Current focus: 1-Hz generative world model driving a 10-Hz policy through nuPlan closed-loop with stable trajectories ≥30 s.
Background
PhD UC San Diego 2023, advised by Hao Su at the SU Lab; thesis on action-conditioned world models for autonomous driving. Visiting researcher at Wayve London (2022) contributing to GAIA-1. Joined Blueberry from Cruise World-Model Team (2023–2025) where she led the closed-loop CARLA + log-replay evaluation pipeline.
Education
- PhD · Computer Science — UC San Diego (2023), advisor: Hao Su
- BS · Computer Science — KAIST (2017)
Selected publications
- Kang et al., 'Action-Conditioned Latent World Models for Urban Driving', CoRL 2022
- Kang et al., 'Closed-Loop Evaluation of Driving Policies via Generative Replay', CVPR 2024
- Kang et al., 'Counterfactual Scenario Generation for Long-Tail AD Failures', ICCV 2025
Reading list
- Hu et al., 'GAIA-1: A Generative World Model for Autonomous Driving', Wayve tech report 2023
- Wang et al., 'DriveDreamer: Towards Real-world-driven World Models for Autonomous Driving', ECCV 2024
- Yang et al., 'Generalized Predictable World Models for Driving (GenAD)', CVPR 2024
- Hafner et al., 'DreamerV3: Mastering Diverse Domains through World Models', Nature 2025
- Caesar et al., 'nuPlan: A closed-loop ML-based planning benchmark for autonomous vehicles', CVPR 2022
Lead experiments
All experiments →No experiments led by this member yet.