Mind the Sim-to-Real Gap & Think Like a Scientist

In three linesTheoretical work on balancing pre-trained simulators with real experiments in sequential decision-making. Decomposes simulator error into calibration-deployment shift and parametric residual. Proposes Fisher-SEP, an experimental policy minimizing posterior predictive variance. Case studies: vending-machine supply chain and HIV mobile testing.

Summary generated by Claude — human-verified