UtilityMax Prompting: A Formal Framework for Multi-Objective Large Language Model Optimization

Ofir Marom · 2026

Read UtilityMax Prompting: A Formal Framework for Multi-Objective Large Language Model Optimization by Ofir Marom (2026) 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-2603-11583.txt.

Description

Marom introduces UtilityMax Prompting, a formal way to state multi-objective tasks for large language models. Plain-language prompts blur what a task wants when several goals must hold at once. The framework rebuilds the task as an influence diagram, with the model's answer as the sole decision variable. A utility function over the diagram's parts then names the target: the answer with highest expected utility, forcing point-by-point reasoning. On MovieLens 1M, the method beats plain prompts on precision and NDCG across three frontier models.

Catalog record

CREATOROfir Marom
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE19979
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.11583

Browse the full library · About the Tower of Babel