Can Large Language Models Solve Engineering Equations? A Systematic Comparison of Direct Prediction and Solver-Assisted Approaches

Sai Varun Kodathala, Rakesh Vunnam · 2026

Read Can Large Language Models Solve Engineering Equations? A Systematic Comparison of Direct Prediction and Solver-Assisted Approaches by Sai Varun Kodathala, Rakesh Vunnam (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-2601-01774.txt.

Description

Engineering runs on equations that no formula solves directly, from pipe friction to orbital position, and practitioners increasingly ask language models to just give the number. Sai Varun Kodathala and Rakesh Vunnam test six leading models on one hundred such problems. The problems span seven engineering domains. Two approaches compete. One asks for direct prediction. The other is a hybrid: the model sets up the equation, and a classical solver iterates.

Direct answers were far off, with mean relative errors around or above one in the worst cases. Letting a Newton-Raphson solver do the arithmetic cut errors by roughly two thirds to four fifths, with the biggest gain in electronics and the smallest in fluid mechanics. The lesson: today's models excel at symbolic setup and recalled knowledge but should hand precision arithmetic to real solvers.

Catalog record

CREATORSai Varun Kodathala, Rakesh Vunnam
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE43 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-01774

Browse the full library · About the Tower of Babel