Test Star released the first vertical large model in the field of test evaluation.

Recently, Exam Star released the first large vertical model in the field of exam evaluation. Before this release, the large model has been tested in several projects. This big model has the characteristics of verticality, scene and high precision, and it integrates language big model, image big model and multi-modal big model, which can help enterprises and institutions to solve recruitment, certification, talent ability evaluation and other scenarios and provide more intelligent and efficient solutions.

2024 is the 10th year of the establishment of Exam Star. In the past 10 years, Exam Star has been focusing on the field of digital examinations. At present, Exam Star platform has 550,000+registered enterprises, delivering 1.57 million online examinations and serving nearly 100 million candidates. In 2021, Exam Star innovatively put forward the concept of "serious examination" in the industry, and realized online examination and online invigilation through pure online mode to ensure the fairness of online examination; In 2023, Exam Star released a number of functions and schemes based on the big model, which further enhanced the efficiency and accuracy of the whole process of various exam scenarios. After half a year of project measurement and data fine-tuning, Exam Star officially released the vertical big model in 2024, and opened it to the public for free testing.

This is the first release of the vertical big model of examination evaluation in the industry, which is another milestone in the field of examination evaluation based on years of industry insight and service experience and embracing the leading big model technology. In the whole process of examination evaluation, the big model can solve the intelligent scheme of proposition, invigilation, grading/evaluation and so on.

In terms of proposition, LLM intelligently sets questions, and through fine-tuning training of 530,000 questions, it has been able to support the requirements of most general scenarios, as well as the requirements of 171 vocational qualification examinations and professional field certification. After the certification project test, using LLM intelligent problem-setting can reduce the problem-setting cost by 95% and improve the problem-setting efficiency by more than 14 times.

In the invigilation scene, multi-modal AI-assisted invigilation is realized through gesture recognition, facial recognition and sound content monitoring, and the cheating tendency in the exam is graded based on a certain algorithm, so that hierarchical and accurate invigilation is realized, which greatly improves the invigilation efficiency of online exams. Based on the cheating database generated by more than 2 million exams, the invigilation can be improved by 10 times.

In the LLM grading scenario, after 11 million test papers’ data training and model fine-tuning, the phase speed compared with manual grading can reach 90%~95% through project measurement, thus achieving a credible and usable state. It can be applied to the scoring of various subjective test papers, and the scoring efficiency can be improved by more than 7 times. In 2024, it will also carry out in-depth cooperation with a number of institutions to open a pilot and benchmark for large-scale LLM scoring in entrepreneurship.

In addition to the general large-scale model technology manufacturers, Exam Star will also cooperate strategically with Tsinghua University on multi-modal large-scale models to promote large-scale models to go deep into the industry. Next, Exam Star will carry out in-depth cooperation with more and more enterprises and institutions on large-scale model business, explore and optimize multiple scenarios in the field of exam evaluation, and promote high-quality development in the field of exam evaluation.

Disclaimer: The market is risky, so you should be careful when choosing! This article is for reference only, not as a basis for trading.