UNIST Department of Industrial Engineering has marked a new milestone in financial AI research, as Professor Yongjae Lee’s team delivered a standout performance at the ACM International Conference on AI in Finance 2025 (ICAIF’25)—presenting the highest number of accepted papers than any other individual research group.
ICAIF, organized by the Association for Computing Machinery (ACM), is widely regarded as one of the leading global conferences at the intersection of AI and finance. It brings together major financial institutions—including JP Morgan and BlackRock—as well as prominent academic researchers from around the world.
Held in Singapore from November 15 to 18, ICAIF’25 accepted only 111 out of 349 submissions (31.8%). Professor Lee’s team presented eight papers—three oral and five poster presentations—outperforming research groups, such as Professor Mihai Cucuringu’s team at UCLA (6 papers) and Dr. Dhagash Mehta’s group at BlackRock (5 papers).
The team’s research also gained significant attention. Their paper, ‘Your AI, Not Your View: The Bias of LLMs in Investment Analysis,’ drew considerable interest from both academia and industry. The team was invited to present their findings at the Bank of Korea and the Korea Investment Corporation (KIC), with additional requests from UBS (Hong Kong), Bloomberg (New York), and several major hedge funds.
The study presents clear evidence that leading large language models—including ChatGPT, Gemini, Llama, and DeepSeek—display distinct investment biases influenced by industry sector, company size, and investment strategy. The researchers also found significant variation in how models adjust their recommendations when confronted with contradictory information, suggesting that training-induced biases do not always adapt smoothly to real-time market conditions.
Professor Lee emphasized the importance of these insights, noting, “As more financial firms adopt LLM-powered tools, understanding the specific biases of each model is essential before deploying them in practice.” The team has released its evaluation as a public leaderboard and plans to expand it as new models become available.
Beyond their research output, the team contributed to the qualitative growth of ICAIF’25. They co-organized three workshops—on LLMs and Generative AI in Finance, Financial Time-Series Forecasting, and Risk Modeling in Emerging Markets—and hosted both an agent-based financial document retrieval competition and a tutorial on decision-focused optimization for finance.
This strong presence reflected a broader rise in Korea’s standing in the field. Korean corresponding authors accounted for 14% of all accepted papers, ranking third globally after the United States (31%) and the United Kingdom (27%). Many attendees noted that UNIST’s contributions significantly elevated Korea’s visibility in the global AI-finance research community.
Professor Lee has steadily expanded his international engagement since first participating in the ICAIF’22, serving on the organizing committee and as a workshop chair. Reflecting on this year’s success, he said, “Returning to ICAIF three years later as the group with the most accepted papers is especially meaningful. I am grateful to the students and colleagues whose hard work made this possible.”
He also expressed hopes for the continued rise of Korea’s AI capabilities, saying, “As Korea’s AI research ecosystem grows, I would like to see world-class conferences like ICAIF hosted in Seoul in the near future.”
Professor Lee currently serves on the Presidential Council on National Artificial Intelligence Strategy, sits on the AI Advisory Council of the Financial Services Commission, and advises global AI-finance firm LinqAlpha, helping to ensure that the team’s research translates into real-world impact. Future ICAIF venues include Milan in 2026, the United States in 2027, and a planned Asia location in 2028.
