- Public call for dataset-construction consortia open from 17 July to 7 August 2025



The Ministry of Science and ICT (MSIT, Minister Bae Kyung-hoon) and the National Information Society Agency (NIA, President Hwang Jong-sung) are openly inviting applications—from 17 July to 7 August 2025—for organizations that will carry out the “Performance Evaluation Dataset Construction Project.”



The project will build the datasets needed to assess the AI models now being developed by the elite teams selected for the “Proprietary AI Foundation Model” initiative, and is therefore a core pillar of that initiative.



Global big-tech companies have released a range of generative-AI services, yet most benchmark tests still rely on English-language metrics and do not adequately reflect the usage environment of Korean-language services.



To address this gap, MSIT will invest KRW 2.4 billion—three tasks at KRW 800 million each—to construct high-quality performance-evaluation datasets that embody Korea’s culture and social values while enabling an objective diagnosis of both domestic and overseas AI models.



After consulting Korean and international experts in AI-model development and evaluation across academia and industry, the ministry identified three priority dataset areas for this year:



1. Mathematics – datasets for gauging large language models’ (LLMs) math problem-solving ability;

2. Korean Knowledge – topic-specific question-answering and reasoning datasets to evaluate Korea-centric knowledge;

3. Long-context Understanding — datasets for testing model performance on tasks that require comprehension of extended passages.



In future phases, datasets will also be developed to evaluate additional generative-AI domains, including multimodal models and AI agents.



Companies or institutions wishing to participate must form a consortium that includes at least one organization with proven capabilities in developing large-scale AI based on extensive datasets—such as hyperscale AI, natural language processing (NLP) models, or multimodal AI.



Kim Kyung-Man, the Director General for the Artificial Intelligence Policy Bureau at MSIT, stated, “For Koreans to reap the full benefits of high-performance, home-grown AI foundation models, we need evaluation datasets that faithfully capture our own social and cultural context. The datasets created through this project will be released openly so that, beyond the elite development teams, any domestic AI organization can use them—ultimately sharpening the competitive edge of Korea’s AI ecosystem.”







For further information, please contact the Public Relations Division (Phone: +82-44-202-4034, E-mail: msitmedia@korea.kr) of the Ministry of Science and ICT.



Please refer to the attached PDF.