LLMPC: Large Language Model Predictive Control

Gabriel Maher · 2025

Read LLMPC: Large Language Model Predictive Control by Gabriel Maher (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-2501-02486.txt.

Description

Gabriel Maher viewed language model prompting through the lens of control theory in a 2025 paper. Model predictive control, a workhorse of engineering, plans by repeatedly minimizing a cost over a short horizon. Prompting tricks have separately made language models better at reasoning, planning, and acting, but the two literatures rarely speak to each other.

Maher argues that, under planning prompts, a language model behaves like an implicit cost minimizer of exactly that kind. He builds a unified framework, LLMPC, that casts prompt design as predictive control design. On several planning benchmarks the framework beat standard few-shot prompting. The bridge may let engineers tune model planners with decades of control insight instead of folklore.

Catalog record

CREATORGabriel Maher
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE38 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-02486

Browse the full library · About the Tower of Babel