SK Telecom announced that a new AI recommendation model that more accurately identifies customers’ true preferences was presented at the International Conference on Learning and Representation (ICLR) 2026, a major AI conference.
ICLR is considered one of the top three AI conferences in the world, along with the Conference on Neural Information Processing Systems (NeurIPS) and the International Conference on Machine Learning (ICML).
The technologies announced at these events will have industrial applications or become benchmarks for academic research and product development within the next few years, providing important indicators of the future direction of AI.
The research paper, titled “More Than What Was Chosen: LLM-Based Explainable Recommendation Beyond Noisy User Preferences,” focuses on SKT’s proprietary ConflictAware Direct Preference Optimization (C-APO) technology. This technology filters out conflicting preference signals, allowing us to more accurately recommend products that customers really want.
To achieve this, SKT introduced the concept of ‘consistent preferences’, which reflect recurring patterns and contextual consistency across a customer’s overall behavior. Through this approach, the model distinguishes between consistent preferences and revealed preferences that derive from a single action, such as an accidental click or a one-time unintentional selection.
In this approach, the learning weight is increased and the recommendation is strengthened when the two preferred signals match. When the signals are inconsistent, the model treats the discrepancies as noise and lowers the learning weights to help identify the most appropriate recommendations.
Traditional AI recommendation systems struggle to interpret conflicting preference signals. For example, a customer who typically watches romance movies may only receive action recommendations, even though they watched one action movie based on a friend’s suggestion. However, SKT’s new recommendation model filters out such discrepancies and identifies the customer’s true preferences.
SKT plans to apply this research to future agent-based customer recommendation systems that integrate AI to interpret usage, generate recommendations, and orchestrate results through an orchestration layer. The company aims to provide more sophisticated and personalized recommendations tailored to the needs and behaviors of each customer.
Seok Ji-hwan, SK Telecom AI/DT data office manager
SK Telecom AI/DT Data Office Head Seok Ji-hwan said this technology lays the foundation for providing personalized recommendations through a deeper understanding of customers’ true preferences. The presentation at ICLR 2026 marks a milestone in SKT’s AI transformation to improve the customer experience, and the company continues to advance AI technology for the benefit of its customers.
