*TITLE*:Benchmarking Large Language Models for Solving
Ordinary Differential Equations

*ABSTRACT*: Solving Ordinary Differential Equations (ODEs)
requires both mathematical modeling and reasoning skills, tradi-
tionally reserved for trained experts. With the rapid advance-
ments of Large Language Models (LLMs), there is growing
interest in evaluating their potential to automate complex math-
ematical tasks, including those involving ODEs. In this paper, we
introduce a novel benchmark for first-order, open-ended ODE
problems and propose an approach for automatically generating
step-by-step solutions, including integration with external numer-
ical solvers. The benchmark evaluates a diverse set of proprietary
and open-source LLMs on a curated dataset of ODE problems.
Using a controlled three-shot prompting strategy, we assess model
performance across several dimensions. Our findings highlight
the current limitations of LLMs in reasoning about open-ended
ODE problems and suggest substantial room for improvement in
this domain