ChatGPT has achieved the highest scores in this year’s entrance examinations for the University of Tokyo and Kyoto University, two of Japan’s most prestigious universities, surpassing the top-performing human candidates, according to an AI venture on Monday, News.Az reports, citing Kyodo.
The AI, developed by LifePrompt Inc., scored 50 points higher than the top human test-taker in the University of Tokyo’s highly competitive Natural Sciences III medical track exam and also achieved a perfect score in mathematics.
The result marks a significant improvement from 2024, when the AI failed to pass all of the university’s entrance exams.
The company conducted the evaluation using OpenAI’s ChatGPT 5.2 Thinking model. It fed the system exam questions converted into image data and assessed essay-type responses with grading assistance from teachers at the major cram school Kawai Juku.
In addition to the University of Tokyo exams, ChatGPT also completed Japan’s unified university entrance examinations, with LifePrompt calculating total scores.
At the University of Tokyo, the AI scored 452 out of 550 on the Humanities and Social Sciences exam and 503 out of 550 on the Natural Sciences exam. Both results exceeded the highest scores recorded among successful applicants, which stood at 434 points for Humanities and Social Sciences III and 453 points for Natural Sciences III.
Performance varied by subject. The AI achieved 90 percent in English but only 25 percent in essay-based sections such as World History.
For Kyoto University, ChatGPT scored 771 points in the Faculty of Law exam, surpassing the top passing score of 734, and 1,176 points in the Faculty of Medicine exam, exceeding the highest human score of 1,098.
In 2024, LifePrompt tested ChatGPT-4 on the University of Tokyo entrance exam, but it failed to reach the minimum passing score. The following year, the upgraded o1 model successfully cleared the threshold for the first time.
“The AI’s capabilities have been well documented. Given the rapid pace of AI evolution, companies will need to adopt AI with an eye toward how business operations will look in 10 to 20 years,” said Satoshi Endo, head of LifePrompt.
Satoshi Kurihara, a professor at Keio University and head of the Japanese Society for Artificial Intelligence, said humans should not be directly compared with AI in such exams because AI is designed to process vast amounts of existing data.
“Just as calculators can perform calculations faster and more accurately than humans can, it is only natural for AI to earn high scores,” Kurihara said, adding that humans remain superior in creating new value. He also suggested that entrance exams should be reconsidered, as they currently emphasize memorization and calculation skills.
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。