• 2 min read
VK algorithm predicts users' future tastes
VK Research presented an algorithm that predicts how recommendations may change users' future interests, with tests underway inside the VK ecosystem.
Image: ITzine
VK Research has developed a recommendation algorithm that looks beyond what users like now. It also estimates how today’s recommendations could shift their interests tomorrow. The approach is already being tested inside the VK ecosystem across several tasks, and the research was presented at the KDD 2026 conference.
Rather than optimizing only for an immediate click, the system evaluates delayed effects. A successful recommendation could keep someone engaged with a service for longer, while a poor suggestion may push the user toward interests the platform will struggle to match later.
The method builds on conventional recommendation systems, which remain useful when speed and low computational costs matter. VK says it can be integrated into an existing architecture without rebuilding the platform from scratch. On public datasets, including VK-LSVD, it produced more accurate content recommendations than traditional methods.
Spotify, YouTube, and major e-commerce platforms have also been testing systems that account for longer-term user behavior rather than one-off interest. For VK, the practical goal is to improve its recommendation products in a competition for attention where feed errors are quickly visible.
Researchers plan to incorporate user feedback more precisely and expand the technology to additional industrial applications. If its accuracy holds on real service data, VK could use the method to tune feed and video recommendations without replacing its existing systems.
Recommended reading
Yandex AI now finds stations with the right fuel
Maxim Tretyakov is a technical journalist who primarily covers the mobile phone market and automotive technology. He has written 740 articles on flagship smartphone launches, including Xiaomi and Apple products, neural-network features in consumer devices, and the state of Russia’s domestic auto industry. His expertise spans hardware releases—from graphics card concepts to smart rings—as well as legal issues in the technology market. Twitter
AI Editor
Ava covers the rapidly evolving world of artificial intelligence, from foundational models and research labs to the real-world economics of intelligence. With a background in computational linguistics, she cuts through the hype to find out what actually works. She firmly believes that benchmarks are just marketing until reproduced in the wild.
via ITzine
