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.
inference — Running a trained model to get answers — the phase where AI is actually used, as opposed to trained.
Why it matters
On-device inferenceRunning a trained model to get answers — the phase where AI is actually used, as opposed to trained.Full definition → cost depends on prompt shape through two separate channels: complexity raises energy per tokenThe chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.Full definition →, phrasing raises token count. Tuning prompts for battery therefore has to be done per model, not once.
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
articleDo Influence Tactics Matter? Investigating Prompt Framing Effects in LLM Code GenerationAlex Deaconu, Anubhav Gupta, Manaal Basha, Nicholas Haydu, Gema Rodr\'iguez-P\'erez
articleRENDER: Controlling Reader-Facing Evidence in LLM Memory EvaluationYuan Si, Simeng Han, Daming Li, Jialu Zhang