Mitihon – SCBX proudly announces its achievement of ranking 5th among 40 teams worldwide, earning the highest score among industry participants in the Retrieval-Augmented Generation (RAG) system development competition at ACM SIGIR 2025, the world’s leading conference on Information Retrieval, held for over 48 consecutive years.
ACM SIGIR is recognized as the premier global academic conference in Information Retrieval, serving as the birthplace of groundbreaking innovations such as Google’s search engine and recommendation systems used by TikTok and other leading digital platforms worldwide.
This year’s competition attracted participants from top universities and global tech giants, including Carnegie Mellon University, New York University, Seoul National University, Amazon Web Services, and several leading startups.
Competition Challenges
The competition posed significant challenges. All participating teams were required to work with the same dataset and language models, making success dependent on unique ideas, methodologies, and technical approaches. The competition comprised three key stages:
- Abstract Submission for peer review
- RAG System Development within a limited timeframe using a database of over 15 million documents
- System Testing with 500 questions in just 2 hours
The SCBX research team delivered an outstanding performance, securing 5th place globally and achieving the highest score among industry teams. They were also honored with both Oral and Poster Presentation slots at this prestigious A* (A-Star) academic conference.
DoTA-RAG: An Intelligent Assistant for Faster, More Accurate Online Information Access
In today’s world of overwhelming online data, finding precise information is no easy task. To address this challenge, SCBX developed DoTA-RAG (Dynamic-of-Thought Aggregation RAG)—a next-generation RAG system designed to enable AI to retrieve external information quickly and accurately.
DoTA-RAG’s key innovation lies in its multi-layered verification process. Instead of immediately using retrieved answers, the system evaluates and ranks the credibility of multiple sources before delivering the final response. It also introduces a new type of database architecture that boosts search efficiency across large datasets, enabling faster, more relevant, and trustworthy access to online information.
From Global Stage to Business Applications
This success is credited to the collaborative efforts of AI researchers and engineers from SCBX R&D and Innovation Lab and SCB 10X, including Saksorn Ruangtanusak, Natthapath Rungseesiripak, Peerawat Rojratchadakorn, Monthol Charattrakool, and Natapong Nitarach, who transformed internal RAG research into a globally recognized solution.
SCBX plans to apply this research to business operations within the SCBX group, such as developing RAG systems for financial document analysis to assess SCB Credit Risk, and leveraging selected techniques for upcoming international hackathons.
ติดตามช่องทางมิติหุ้นเพื่อรับข่าวสารตลาดทุนได้ตามลิงค์ด้านล่าง
Web : https://www.mitihoon.com/
Facebook : https://www.facebook.com/mitihoon
Youtube : https://www.youtube.com/@mitihoonofficial7770
Tiktok : www.tiktok.com/@mitihoon
