Simulation Distillation (SimDist) enables rapid sim-to-real adaptation by transferring reward and value models directly into a latent world model.
March 18, 2026
Original Paper
Simulation Distillation: Pretraining World Models in Simulation for Rapid Real-World Adaptation
arXiv · 2603.15759
The Takeaway
It converts the difficult problem of real-world long-horizon reinforcement learning into a simpler short-horizon system identification task, greatly increasing data efficiency and stability for robot learning.
From the abstract
Simulation-to-real transfer remains a central challenge in robotics, as mismatches between simulated and real-world dynamics often lead to failures. While reinforcement learning offers a principled mechanism for adaptation, existing sim-to-real finetuning methods struggle with exploration and long-horizon credit assignment in the low-data regimes typical of real-world robotics. We introduce Simulation Distillation (SimDist), a sim-to-real framework that distills structural priors from a simulato