Seoul National University will host a large-scale artificial intelligence (AI) competition for university students nationwide, offering the president's award and a total prize pool of 23 million won. The Graduate School of Data Science at Seoul National University said on the 3rd that it will hold the "SNU AI Challenge 2026," themed "reconstructing scenes through text."
The competition is organized by the Graduate School of Data Science at Seoul National University and sponsored by More, Motif Technologies, the Center for Optimization of Foundation AI Models and Platforms, and the BK21 Data Science Talent Development Program.
This competition is a high-difficulty technical task in Multimodal AI that requires inferring temporal context by using scene information together with textual clues. Participants will develop AI models that reconstruct randomly shuffled scenes into the correct order according to given text clues. A dedicated dataset, specially processed by the graduate school for the competition, will also be released.
In particular, rather than competing solely on accuracy, the competition requires completing inference within 24 hours using a single NVIDIA RTX 3090 GPU (VRAM 24GB). The aim is to comprehensively evaluate model performance, efficiency, and practicality within limited computing resources and time.
Participants may enter individually or in teams of up to four. The online preliminaries will run from June 29 to July 24, and applications are due by July 17. Teams that pass the preliminaries will advance to the finals, where they will be selected as the winners after a public presentation evaluation on Aug. 7. The total prize pool is 23 million won, including the grand prize (10 million won), and outstanding teams will receive the Seoul National University president's award.
Yu Hong-lim, president of Seoul National University, said, "In the AI era, universities should not merely follow technology but reflect on and shape the direction AI should take based on human-centered values," and added, "I hope this competition becomes an opportunity for students to take the lead in the entire AI development process, from problem definition and data understanding to model design and optimization, system implementation, and performance verification."
Lee Jae-jin, dean of the Graduate School of Data Science, said, "It is meaningful to build practical AI capabilities, as we consider model performance, efficiency, and real-world applicability under constrained resources," and noted, "We will support students through convergent education and research that connect data, algorithms, systems, and domain knowledge."
Applications and details are available on the SNU AI Challenge website at SNU AI Challenge website.