Incorporating Token Usage into Prompting Strategy Evaluation

Chris Sypherd, Sergei Petrov, Sonny George, Vaishak Belle · 2025

Read Incorporating Token Usage into Prompting Strategy Evaluation by Chris Sypherd, Sergei Petrov, Sonny George, Vaishak Belle (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-14880.txt.

Description

Chris Sypherd, Sergei Petrov, Sonny George, and Vaishak Belle argued in 2025 that prompting strategies for language models should be judged on token cost as well as accuracy. A clever prompt can win on score while burning many more tokens, which costs money and time. The usual score-only ranking hides that trade-off.

The team proposes a theory frame that describes how token use grows with each strategy, in the style of big-O notation, plus a hands-on measure of tokens spent per unit of performance. Applied to several common strategies, both tools show that extra tokens bring steeply shrinking gains. Readers get a practical correction to prompt engineering culture: count the tokens, not just the wins, before calling a strategy better.

Catalog record

CREATORChris Sypherd, Sergei Petrov, Sonny George, Vaishak Belle
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE48 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-14880

Browse the full library · About the Tower of Babel