HSE students successfully defended diplomas, in the work on which YandexGPT was used

2024

Degree protection, in the work on which YandexGPT was used

In the summer of 2024, students in several areas of training at the Higher School of Economics - Philosophy, Media Communications, International Relations and Oriental Studies - first used the neural network in the process of writing thesis and course papers. The university announced this on August 7, 2024.

The university allowed the use of the capabilities of YandexGPT, Yandex's generative technology, to solve a predetermined list of tasks that usually take a student a long time. She helped collect, analyze and summarize information, check text for errors and edit the design, as well as understand complex topics and structure the course of reflection.

More than half of the students who applied to AI during preparation received the highest score. This is the first such experience in the system integration of AI technology of bigtech into the educational process. The HSE considers the experiment successful and from 2025 will increase the number of areas where YandexGPT can be used.

Obviously, generative models are used everywhere - including by students when performing certain tasks. I am sure that in terms of design and research work, it is worth encouraging the use of large language models. It is important for us that the student gets a successful result or creates a new product, and if the AI tools helped him in this, accelerated the work and improved its quality, we are only in favor. Therefore, we at the Higher School of Economics generally allow the use of generative tools in thesis works, and also gladly took part in a pilot project with Yandex, "said Evgeny Sokolov, head of the department of big data and information search at the Higher School of Economics.

As part of the project, students learned to competently use generative technologies and were able to dive deeper into research, reducing the time for routine tasks and freeing them up for creative ones. In their works, students noted exactly where they used neural network tools, and on the defenders they told in more detail what tasks they turned to the AI assistant and how they carried out factchecking.

According to polls, the most popular function was the search for errors and typos, in second place was the summarization of data. In addition, students used the neural network to discuss various aspects of the work, turned to it to generate images and for help in design. At the same time, students independently drew conclusions in their diploma projects. The quality of the work done and how deeply the student is immersed in the topic were checked during the defense before the commission.

Once upon a time, the advent of the Internet accelerated the search for information and changed the educational process. Neural networks help students and teachers better navigate the flow of information. Our joint project with HSE is a step towards the meaningful use of AI technologies in the educational process. This experience will be useful for students in further work, where the skills of using AI tools and critical thinking are becoming more and more in demand, regardless of the field. Our task is to provide conditions and give access to technologies that are being created right now, "said Darya Kozlova, director of Yandex Education.

In addition to conducting research and successfully defending diplomas, students got an idea of ​ ​ how future professional activities can work in tandem with AI assistants and tools.

Using YandexGPT

The National Research University Higher School of Economics (HSE) tested the solution using YandexGPT, a generative language model available on the Yandex Cloud platform . A chat bot on the university's website and telegram channel will help applicants choose one of several hundred educational programs based on their experience, personal preferences, course complexity and other parameters. Everyone from June 2024 will be able to use the service on the HSE website. The university announced this on May 24, 2024.

The chatbot was developed by HSE specialists and Cloud architects Yandex. For additional training of the model, the RAG (Retrieval Augmented Generation) method was used - an approach in which neuronets an additional information from external sources in the form of context is transmitted to answer the question. The technology works like this: the user in free form tells the bot about himself and his interests. The neural network analyzes the information and offers the applicant a suitable program.

Every day, more than 7 thousand people go to the main page of our site and look for information there about which educational program to choose for them. The presented bot using generative models will help facilitate such a choice and get a balanced answer to the question "Who will I become when I grow up," said Dmitry Koptyubenko, director of the HSE portal.

In the future, the university plans to use the YandexGPT API service in other tasks - for example, to recommend students to "tighten" knowledge of a particular subject in additional courses or give advice on the topics of graduation qualifications.

The HSE also began using machine learning technologies in internal processes to automate the routine tasks of university employees. For example, a neural network based on Yandex Vision OCR and YandexGPT API checks the authenticity of certificates of temporary disability of students and will then fill out data about them in the university's information system. Potentially, with the help of the system, it will be possible to process up to 20 thousand such documents per year.