SK Telecom (SKT) is collaborating with domestic mathematics and artificial intelligence (AI) researchers to enhance the mathematical reasoning capabilities of its proprietary AI foundation model, A.X K2.
On October 22, SKT announced that it would hold a technology exchange meeting at its headquarters in Euljiro, Seoul, with professors and researchers from Seoul National University and KAIST to discuss research on mathematical challenges and strategies for model advancement using A.X K2.
The project involves researchers from the fields of mathematics and AI. Notable participants include Professor Seo In-seok from the Department of Mathematical Sciences at Seoul National University, who specializes in probability and statistical mechanics; Professor Lee Jun-kyung from the Institute for Basic Science, who focuses on discrete mathematics and graph theory; and Professor Kim Sang-hyun from the Institute for Basic Science, who studies low-dimensional topology and geometric group theory.
The researchers will also explore methodologies for applying large language models (LLMs) to mathematical research. Professors Hong Young-jun from Seoul National University’s Department of Mathematical Sciences and AI program, along with Professors Lee Da-bin, Oh Byeong-kwon, and Byeon Seong-su from the same department, and Professor Park No-sung from KAIST’s Department of Computer Science, will seek ways to incorporate mathematicians' expertise into model training.
The team plans to supplement A.X K2 with specialized knowledge and reasoning abilities in mathematical subfields that existing AI models have not adequately learned. They will then apply the enhanced model to various mathematical research areas to test its potential for solving complex problems.
SKT aims to leverage the mathematical reasoning and verification capabilities gained through this academic collaboration to further develop its agentic AI. The goal is for AI to autonomously assess the validity of its judgments and outcomes while devising solutions to complex problems.
In the long term, SKT plans to expand its research into recursive self-improvement (RSI), allowing AI to enhance its capabilities and utilize them for future advancements.
Kim Tae-yoon, head of SKT's foundation model division, stated, "We will enhance agentic AI capabilities through collaborative research with expert researchers."
Meanwhile, SKT has achieved first place in the third phase evaluation of its foundation model.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.