The participants had to imagine that they are building an assistant bot for architectural and construction firms. Such a bot needs to be able to quickly find the relevant standard within a vast body of documents (SNiPs, SPs, GOSTs, Federal Laws) and produce a clear, coherent answer to a specialist's question.
The task was as follows: given a question, participants had to find the relevant fragment of a regulatory document and generate a detailed answer as close as possible to the reference answer. In essence: to teach the model to "read" complex technical standards and correctly explain what they say.
In the evening, a detailed review of the solutions was held, led by Albina Sergeevna Burlova — Senior Lecturer at the Department of Big Data and Information Retrieval, Faculty of Computer Science, HSE University; Research Intern at the Research and Study Laboratory of Computational Pragmatics, Faculty of Computer Science, HSE University; and Methodologist at the Center for Continuing Education, Faculty of Computer Science, HSE University.
The task was as follows: given a question, participants had to find the relevant fragment of a regulatory document and generate a detailed answer as close as possible to the reference answer. In essence: to teach the model to "read" complex technical standards and correctly explain what they say.
In the evening, a detailed review of the solutions was held, led by Albina Sergeevna Burlova — Senior Lecturer at the Department of Big Data and Information Retrieval, Faculty of Computer Science, HSE University; Research Intern at the Research and Study Laboratory of Computational Pragmatics, Faculty of Computer Science, HSE University; and Methodologist at the Center for Continuing Education, Faculty of Computer Science, HSE University.