Large Language Models exhibit limited reasoning ability on coding questions (submitted)