Slides from my two-part lecture at the ELLIS Summer School on Autonomous Driving, Barcelona 2026.
How offline RL regularizations interfere with online training.
Why it works so well and where the limitations are.
