Reasoning, Code, or Both? How Large Language Models Handle Variations in Math Questions

Large language models (LLMs) show reduced accuracy on math problems when simple details like names or numbers are changed, and a new study finds that using code execution methods does not improve this robustness. Researchers tested three approaches—chain-of-thought reasoning, single-shot code execution, and iterative code execution—on 1,000 modified problems from the GSM-Symbolic dataset using Claude Haiku 4.5. Chain-of-thought proved the most robust, with a 1.3 percentage point accuracy drop and 1.8% of problems breaking, while code execution methods performed worse, indicating that generating and running Python code does not enhance reasoning stability for grade-school-level math variations.

arXiv:2605.26414v1 Announce Type: new Abstract: Large Language Models LLMs achieve impressive accuracy on mathematical reasoning benchmarks, yet their performance drops when problems are modified with simple changes like different names or numbers. Code execution methods, which let models generate and run Python code instead of reasoning in natural language, have been proposed as a solution, but their effect on reasoning robustness the ability to maintain accuracy across problem variations has not been systematically tested. This study evaluates three approaches on 1,000 problems from the GSM-Symbolic dataset: pure reasoning using chain-of-thought CoT prompting, single-shot code execution using Program-Aided Language models PAL , and iterative code execution using Step-by-Step Coding SBSC . All three were run on paired original and modified problems using Claude Haiku 4.5. CoT was the most robust method, with an accuracy drop of 1.3 percentage points and 1.8% of problems breaking under perturbation. PAL was the least robust at 1.7 percentage points and 3.1% broke, with SBSC falling in between. Although these differences were not statistically significant $p = .096$ , the directional trend was consistent across all measures, suggesting that code execution, whether single-shot or iterative, does not improve reasoning robustness on grade-school-level problem variations.