As large language models (LLMs) are increasingly deployed on mobile and embedded devices to enhance privacy and reduce network latency, on-device inference faces a fundamental constraint: high energy consumption on battery-powered, resource-limited hardware. While model compression and runtime acceleration have been widely studied, the impact of \textit{prompt design} on energy efficiency remains underexplored. This paper presents an empirical study of the relationship between prompt wording and energy consumption for on-device LLMs.
By utilizing real power measurements collected from a smartphone, we quantify how linguistic features, particularly imperative keywords and instruction structure, affect decoding length and total energy. Our results indicate consistent energy differences across verbs and tasks, suggesting that prompt engineering is a lightweight lever for improving energy efficiency.
Blogger's Review: This study highlights the potential of prompt design in enhancing LLM energy efficiency, especially in resource-constrained environments. By optimizing prompt content, significant energy savings can be achieved, which has important practical implications.