Exploring the Robustness of Large Language Models for Solving Programming Problems