Dunamu achieves global AI research results…applies AI-based services within platform
Summary
- Dunamu said it presented papers at global conferences and ranked first on a benchmark leaderboard with research achievements in AI and machine learning.
- It said service innovations were implemented, including the introduction of Upbit Assistant within the platform and the advancement of AI-based investment information services.
- Dunamu said it will continue to strengthen technological competitiveness and provide more reliable services next year.

Dunamu, the operator of the digital asset exchange Upbit, said on the 18th that it strengthened its technological competitiveness this year by achieving global research results in the field of artificial intelligence (AI) and advancing its services simultaneously.
Dunamu's machine learning (ML) team presented a research paper at the main conference of COLING 2025, the International Conference on Computational Linguistics, held in Abu Dhabi, United Arab Emirates, last January. The presented paper was evaluated as having improved response inconsistency issues occurring in LLM-based text-to-SQL conversion through multi-prompting and multiple-choice methods, and it ranked first on the global benchmark BIRD-SQL leaderboard.
Also, Dunamu's ML team presented a paper at SIGIR 2025, the leading international conference in the field of information retrieval, held in Padua, Italy, last July. The paper proposed a method of generating virtual users using LLMs to train news recommendation models without real user data. Dunamu's 'LAUS (LLM As User Simulator)' framework showed superior performance compared to zero-shot methods and also reduced response latency.
Dunamu expanded the application of AI technology to actual services. The Upbit Developer Center introduced an AI-based 'Upbit Assistant', improving developer convenience by providing API examples, document summarization, guide navigation, and more. It enabled major AI tools such as GPT and Claude to use the latest Upbit API documentation, and added the 'Upbit Historical Market Data' service to download past price data, strengthening the data-driven development environment.
Dunamu's investment information platform Securities Plus also enhanced its AI features. The newly established 'Newsroom' is a representative example. Dunamu's Newsroom uses an in-house developed machine learning algorithm to prioritize and provide users with information necessary for investment decisions. Specifically, the 'Breaking' section quickly delivers key points with a headline and a three-line summary, and the 'Major News' section selects and provides the 'Top 5 core news in 24 hours' that had the greatest impact during the day.
Lee Dong-jun, head of Dunamu's ML team, said, "Based on the AI research achievements and service application experience accumulated this year, we will continue to strengthen technological competitiveness next year and provide users with more reliable services."

Bloomingbit Newsroom
news@bloomingbit.ioFor news reports, news@bloomingbit.io

!['Hawkish hold?'…FOMC faces more complex calculus amid Iran war [Analysis+]](https://media.bloomingbit.io/PROD/news/13cd35e7-eb6f-4cff-b93b-70a5c4d2a1de.webp?w=250)



