SK Telecom (017670.KS) has unveiled an artificial intelligence recommendation technology that precisely identifies users' "true preferences" at an international academic conference.
SK Telecom said Friday that it presented a paper on a new recommendation model at ICLR 2026, one of the most prestigious conferences in the AI field.
The core of the paper is the "Conflict-Aware Preference Optimization" (C-APO) technology independently developed by SK Telecom. It is designed to precisely distinguish conflicting preferences mixed within user data, producing results that most closely match a user's actual tastes.
The AI technology analyzes users' long-term behavioral data and defines recurring patterns as consistent preferences. Histories of continued use of specific content or services are regarded as genuine interests, while one-time clicks or incidental consumption are classified as surface-level preferences, with their recommendation weights lowered accordingly.
The technology is seen as improving the recommendation distortion that occurs when existing systems reflect even short-term behavior as user preferences. For example, if a user who typically enjoys the romance genre watches a single action movie on a friend's recommendation, the technology can prevent subsequent recommendation lists from suddenly being flooded with action titles.
SK Telecom's AI technology has also strengthened the logical rationale behind recommendation results. It is expected to contribute to enhancing explainability and reliability in future AI recommendation services, going beyond simply providing results. SK Telecom plans to fully apply this achievement to a three-stage agent-based customer recommendation system that encompasses interpreting usage context, generating recommendation candidates, and verifying and adjusting results.
"This technology is a foundational technology for understanding customers' actual tastes more accurately and providing a more reliable personalized recommendation experience," said Seok Ji-hwan, head of AI/DT Data at SKT. "We will continue to develop AI technologies that customers can tangibly experience."
