Tao Zhexuan actual test o1: it's like giving advice to a mediocre but marginally competent graduate student

Tao Zhexuan tested OpenAI's o1 model and found that it can effectively identify and solve complex mathematical problems, with performance similar to that of a junior graduate student; although the o1 model showed competence in generating hypotheses and solving problems, it still suffers from errors and comprehension limitations, and needs further guidance and tool support; Tao Zhexuan's tests showed that, despite the progress of the o1 model, it still needs to improve in autonomously generating key concepts and error avoidance, and its usefulness and accuracy need to be improved.