A Mechanistic Analysis of Sim-and-Real Co-Training in Generative Robot Policies

arXiv:2604.13645v1 Announce Type: cross Abstract: Co-training, which combines limited in-domain real-world data with abundant surrogate data such as simulation or cross-embodiment robot data, is widely used for training generative robot policies. Despite its empirical success, the mechanisms that determine when and why co-training is effective remain poorly understood. We investigate the mechanism of sim-and-real co-training through theoretical analysis and empirical study, and identify two intrinsic effects governing performance. The first, extbf{structured representation alignment"}, reflects a balance between cross-domain representation alignment and domain discernibility, and plays a primary role in downstream performance. The second, the extbf{importance reweighting effect"}, arises from domain-dependent modulation of action weighting and operates at a secondary level. We validate these effects with controlled experiments on a toy model and extensive sim-and-sim and sim-and-real robot manipulation experiments. Our analysis offers a unified interpretation of recent co-training techniques and motivates a simple method that consistently improves upon prior approaches. More broadly, our aim is to examine the inner workings of co-training and to facilitate research in this direction.

Related

• Generative Simulation for Policy Learning in Physical Human-Robot Interaction
• ComSim: Building Scalable Real-World Robot Data Generation via Compositional Simulation
• IGen: Scalable Data Generation for Robot Learning from Open-World Images
• Sim-to-Real Transfer for Muscle-Actuated Robots via Generalized Actuator Networks
• WM-DAgger: Enabling Efficient Data Aggregation for Imitation Learning with World Models

Source: arXiv cs.LG | 2026-04-16