Over 200 International Students at Tongji University Undertake ‘Foreigner College Entrance Exam’ in China
International students at Tongji University in Shanghai participate in the yanggaokaoor, a key exam for continuing studies in China. Meanwhile, AI models perform well in language sections but struggle with mathematics, emphasizing the importance of balanced development in education.
Over the past weekend, over 200 international students from Tongji University’s International School in Shanghai undertook the yanggaokaoor, also known as the “foreigner college entrance exam.” This annual exam, which assesses students’ progress through subjects such as Chinese language, mathematics, physics, and chemistry, is a key requirement for students to continue their studies in China on a government scholarship supported by the Ministry of Education.
Students faced the exam in Mandarin, with all questions delivered on computers, mimicking the format of China’s national college entrance exams, or gaokao. Preparatory activities by the school’s faculty aimed at alleviating stress echoed those associated with the traditional gaokao.
Guncha Atabayeva from Turkmenistan described the exam as particularly challenging, highlighting the difficulty of taking the test in Chinese. Atabayeva plans to pursue gynecology at Guangzhou Medical University. Claire Maria Kanga Moukendi from the Republic of Congo expressed confidence post-exam, attributing her preparation to the support from her teachers.
Since its inception in 2009, Tongji University’s preparatory program has attracted over 3,300 international students from more than 100 countries, providing a pathway for higher education in China.
Separately, findings from the Shanghai Artificial Intelligence Laboratory revealed that AI models, including GPT-4o, performed well in the Chinese literature and English language sections of this year’s gaokao, scoring 67% and 81% accuracy, respectively. However, they struggled with mathematics, achieving only a 36% accuracy rate.
Alibaba’s Qwen2-72B emerged as the top-performing model with approximately 72% accuracy. Despite their strong performance in languages, AI models faced challenges in understanding classical Chinese text and exhibited disorganized methods in solving math problems, showing a reliance on memorization over problem-solving skills. Researchers and educators underscored the findings, noting the need for balanced development in AI training across various subjects and emphasizing the importance of fostering innovation and critical thinking in human students.