Dunamu’s in-house Machine Learning (ML) team has claimed first place in the Financial Misinformation Detection (FMD) challenge at the International Conference on Computational Linguistics (COLING 2025), the company announced on the 23rd.
COLING is a leading global conference in the fields of Natural Language Processing (NLP) and Computational Linguistics. It brings together researchers and companies to present the latest advancements in technology. The FMD challenge, held as part of COLING 2025, focuses on detecting and analyzing misinformation in financial data. The challenge was created to tackle the growing problem of financial misinformation spreading through media platforms.
Dunamu’s ML team outperformed competitors in key evaluation metrics, including F1 (which measures detection accuracy) and ROUGE (which assesses the reasoning behind misinformation identification). The team achieved this by leveraging Large Language Models (LLMs). Notably, they modified open-source LLMs instead of relying on commercial models like ChatGPT, showcasing their technical expertise.
Hee-Soo Park, a researcher at Dunamu’s ML team, expressed pride in the team’s accomplishment. “Our team has worked tirelessly to protect investors and promote a transparent financial environment. This recognition on an international stage is a testament to our dedication,” Park said.
Dong-Jun Lee, head of Dunamu’s ML team, emphasized the significance of the achievement. “This victory reflects our commitment to building customer trust. We will continue to develop reliable technologies and services that contribute to market stability and transparency,” Lee stated.
The win highlights Dunamu’s leadership in applying advanced machine learning techniques to address real-world challenges in the financial sector.
