The VK RecSys Challenge has concluded, showcasing a remarkable competition in the field of recommender algorithms. Nearly 800 researchers from 14 different countries submitted approximately 3,900 solutions to address the "cold start" problem, which involves predicting user interest in new music videos that have not yet been viewed.
Participants utilized the VK-LSVD dataset, which consisted of 40 billion anonymized interactions with 20 million short videos. The event attracted 219 teams, including many students from prestigious universities such as HSE, ITMO, Moscow State University, and MIPT, who competed in a dedicated student section. The majority of submissions came from Russia, Kazakhstan, Belarus, and Uzbekistan.
The winners, who will share a prize pool of 2.5 million rubles, have already been selected. They are set to be honored at the annual Data Yolka event in Moscow and St. Petersburg in January 2026, which focuses on experts in machine learning and data science. Last year's competition saw around 1,000 participants tackling a similar challenge related to predicting user reactions to video content.
This competition highlights the growing interest and innovation in the field of machine learning and recommender systems, signaling a competitive landscape for tech companies aiming to enhance user engagement through advanced algorithms.
Informational material. 18+.
