Paper Overview
Field: Machine Learning Authors: Angel Wang, Dominique Perrault-Joncas, Alvaro Maggiar, Dean Foster, Carson Eisenach Published: 2026-10-02 arXiv: 2610.03662
Summary
Deploying a new decision policy creates a cold-start problem for prediction models whose targets depend on the policy's actions: historical observations reflect earlier policies, while real observations under the new policy are not yet available. Simulation offers a way to address this gap by rolling out the target policy across counterfactual scenarios and using the resulting trajectories to learn how the system responds to those controls. The simulation-to-reality (Sim2Real) transfer of this simulator-trained model can then be backtested by evaluating it against real observations from past deployments.
Using two real-world inventory-control deployments, the authors evaluate this process from three angles: simulator fidelity, zero-shot transfer to real behavior, and adaptation as real target-policy observations accumulate.
Key Findings
- Simulator-trained predictors achieve lower point-estimate MAPE than the same architecture trained on historical real data:
- Study 1: 1.2–3.1 percentage points lower
- Study 2: 12.5–18.7 percentage points lower
- After deployment, lightweight calibration using early real observations further reduces error by up to 2.5 percentage points.
- These results provide empirical evidence that simulator-generated counterfactual data can support cold-start prediction under new policies, and that the resulting models can be further refined as real deployment data accumulates.
Original Abstract (excerpt)
> Deploying a new decision policy creates a cold-start problem for prediction models whose targets depend on the policy's actions: historical observations reflect earlier policies, while real observations under the new policy are not yet available. Simulation offers a way to address this gap by rolling out the target policy across counterfactual scenarios and using the resulting trajectories to learn how the system responds to those controls. The simulation-to-reality (Sim2Real) transfer of this simulator-trained model can then be backtested by evaluating it against real observations from past deployments. Using two real-world inventory-control deployments, we evaluate this process from three angles: simulator fidelity, zero-shot transfer to real behavior, and adaptation as real target-polic...
*Auto-collected on 2026-10-06*