How a request is worded can change the energy consumed by an AI model running on a mobile device. This study separates energy spent processing the prompt from energy spent generating the response. It finds that phrasing changes mainly affect how many text units the model processes or produces, while changes in reasoning demands also affect the measured energy per unit.
The best balance between answer quality and energy use varies by model. For developers building battery-powered assistants, the finding makes prompt choice something to measure alongside model choice: a wording change that improves one model's efficiency may produce a different trade-off in another.
Terms in this piece · Glossary
token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
articleBeyond Prompt Engineering: A Systematic Analysis of Prompt Lexical Sensitivity and Its Impacts on QualityQipeng Xie, Zi Liang, Jiafei Wu, Yufei Chen, Weizheng Wang, Wenao Ma, Zhong Ming, Haiqin Yang, Kaishun Wu
articleImproving the Energy-Efficiency of the Code Generated by LLMs through Effective PromptingRitika Rekhi, Bing Zhang, Md Arman Islam, Jaya Krishna Pasham, Yeswanth Chitturi, Akshay Paramesha, Isha Valiveti, Asif Imran, Bekir Turkkan, Tevfik Kosar
articleDo Influence Tactics Matter? Investigating Prompt Framing Effects in LLM Code GenerationAlex Deaconu, Anubhav Gupta, Manaal Basha, Nicholas Haydu, Gema Rodr\'iguez-P\'erez