Can Large Vision-Language Models understand graphs? / hacker news