SK Telecom's 'Thinking AI Recommendation Model' Ranks in Top 4% of Papers at Global Conference

On-Site Presentation of 'One Model 4.0' at AAAI 2026

On the 24th, Park Jeong, Head of Data Analysis and Modeling Team at SK Telecom, gave an oral presentation on the ‘One Model’ version 4.0 research at ‘AAAI 2026’ held in Singapore. Photo by SK Telecom

SK Telecom has taken the stage at one of the world’s leading AI conferences with its research on an artificial intelligence (AI) recommendation model capable of understanding individual behavioral contexts and explaining the reasons behind its recommendations.

SK Telecom announced on the 25th that its research paper on its AI recommendation model was invited and presented in person the previous day at the global AI conference AAAI 2026 (Association for the Advancement of Artificial Intelligence), currently being held at the Singapore Expo.

AAAI is recognized as one of the world’s most prestigious AI conferences, alongside NeurIPS (Conference on Neural Information Processing Systems), ICML (International Conference on Machine Learning), and ICLR (International Conference on Learning Representations). This year, about 24,000 papers were submitted to AAAI, a significant increase from approximately 13,000 the previous year. The acceptance rate dropped to 18% from last year’s 23%, making the competition even fiercer.

SK Telecom’s paper was selected for the ‘Oral Session,’ which is reserved for the top 4% of all submissions, and was also presented in the ‘Poster Session’ simultaneously.

The research presented this time focused on version 4.0 of ‘One Model,’ SK Telecom’s proprietary LLM (Large Language Model)-based AI recommendation model. In particular, it provided a concrete example of applying ‘reinforcement learning for reasoning ability,’ a topic gaining attention in global AI research, to a real-world service environment.

One Model 4.0 comprehensively analyzes various behavioral data such as customers’ click histories, usage patterns, and interests to generate in natural language which products or services to recommend, the reasons for these recommendations, and how to compose marketing messages to customers.

This model is notable not only for its research phase but also for its successful application in actual services. It is currently being used for rate plan recommendations, as well as in T World, T Membership, and T Direct Shop, with customer response metrics such as click-through rates improving by up to two times compared to previous versions. SK Telecom plans to expand its application to services like T Universe within the year.

Seok Jihwan, Head of AT/DT Data at SK Telecom, stated, “Through the One Model research, we have once again demonstrated SK Telecom’s AI capabilities by being recognized at globally renowned conferences. We will continue to pursue practical AI research that leads to real products and services.”