The University of Hawaiʻi at Mānoa took first place and a $3,000 cash prize in an international artificial intelligence (AI) competition, beating out teams from tech companies and global universities.
Seven haumāna (students) from the UH Mānoa College of Engineering’s ECE (Electrical and Computer Engineering) 605 course on large-scale AI won the Data Curation for Vision Language Reasoning challenge. The contest drew 59 teams and nearly 200 participants from 42 organizations.
The challenge asked teams to create up to 10,000 high-quality examples to help AI models better understand and reason about images. UH Mānoa’s winning entry stood out not for its size, but for its efficiency. The students used only 1,000 samples—just 10% of the allowed amount—yet outperformed every other team.
“This project pushed us to think creatively about how to get the most out of a tiny amount of data,” team leader Yosub Shin said. “We learned quickly that good ideas matter just as much as computing power, and it was exciting to see our approach stand up against teams from around the world.”
The team’s dataset boosted a widely used open-source model’s accuracy by more than 7%, the highest gain in the competition. It surpassed submissions from industry heavyweights, including a team from Ant Group, the parent company of Alibaba.
Preparing for a fast-growing future
Beyond the recognition, the experience gave students real-world practice in one of the fastest-growing areas of AI: teaching computers how to interpret both language and images. The hands-on training—curating data, testing models, and working within open-source frameworks—prepares them for future roles in fields ranging from robotics to medical imaging.
Members of the team included Shin, Michael Buriek, Boris Sobolev, Pavel Bushuyeu, Vikas Kumar, Haoyang Xu and Samuel Watson.
“Watching the students take ownership of a complex, open-ended challenge like this was remarkable,” ECE Assistant Professor Igor Molybog said. “They showed that with curiosity, discipline, and a willingness to experiment, students can make meaningful contributions to the broader AI research community.”
As the top prize winner, the UH Mānoa team will present their work on stage at the NeurIPS conference in San Diego this December and be listed as co-authors on the competition’s official publication.
Hawaiʻi AI Initiative
The ECE 605 course and the students’ participation in the challenge are part of a larger effort Molybog is driving to promote AI in the state, called the Hawaiʻi AI Initiative (HawAII). The initiative is developing new project-based AI coursework, strengthening internal innovation by supporting student and faculty ventures, and attracting top collaborators from the continental U.S. and abroad through global research challenges. It also includes major investments in UH’s computing infrastructure and early partnerships with local schools to build a strong pipeline of homegrown talent.
