
LLMs exhibit varying levels of math reasoning abilities but lack robustness, especially when faced with question variations.


coremsg

gsm-plus-a-comprehensive-benchmark-for-evaluating-llms-in-math-problem-solving


GSM-PLUS: A Comprehensive Benchmark for Evaluating LLMs in Math Problem Solving



LLMs exhibit varying levels of math reasoning abilities but lack robustness in solving math word problems.



GSM-PLUS: Evaluating LLMs' Robustness in Math Problem Solving
