The Accuracy-Efficiency Paradox Quantifying Net Energy Loss in on-Device Energy Forecasting

Abstract

arXiv:2608.26134v1 Announce Type: new Abstract: Energy forecasting aims to maximize accuracy to ensure energy efficiency by reducing energy waste, an objective that applies equally to on-device forecasting for mission-critical edge environments, including military systems. However, this paper identifies the Accuracy-Efficiency Paradox: high-precision energy forecasting models can ironically trigger a net energy deficit. This stems from both edge AI's inference energy consumption and battery agin