AI RESEARCH
Benchmarking Sensor-Fault Robustness in Forecasting
arXiv CS.LG
•
ArXi:2605.10822v1 Announce Type: new Cyber-physical system (CPS) forecasting models depend on sensor streams with noisy, biased, missing, or temporally misaligned readings, yet standard forecasting evaluation often selects models by nominal error without showing whether they remain robust under such faults. We