AI BriefingKakaoBank Looks Back on if(kakaoAI)2024 ☕️
Key point
KakaoBank introduces key technical sessions presented at if(kakaoAI)2024, including an LLM evaluation framework and XAI acceleration.
Details
KakaoBank participated in a total of 5 technical sessions and 2 panel talks at the if(kakaoAI)2024 conference, sharing its AI technology application cases.
One of the key sessions, DUO, is an LLM evaluation framework specialized for the financial sector, which verifies models based on Financial Calculation Accuracy (FCA) and Financial Multi-turn conversation ability (FMT). This embodies KakaoBank's internalized experience in overcoming the limitations of existing benchmarks and selecting the optimal LLM.
In addition, the results of XAI (Explainable AI) acceleration research were also unveiled. Through industry-academia collaboration with KAIST, an algorithm was developed that improves the speed of the existing KernelSHAP by up to 10x, laying the foundation for applying this to financial services where real-time performance is critical, such as fraud transaction detection.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Designing a Collateral Review System Where People Don't WaitOrganizing Complex Business Logic into Pipelines