VK-LSVD: A Large-Scale Industrial Dataset for Short-Video Recommendation

TL;DR

VK-LSVD tackles the shortage of large-scale, real-world datasets for short-video recommender research by publicly releasing VK-LSVD, which contains over 40 billion interactions from 10 million users and about 20 million videos over six months, with content embeddings and multi-modal feedback signals. The dataset implements a Global Temporal Split and provides rich contextual metadata and anonymized identifiers, enabling robust sequential and cold-start research and custom subset generation. Its value is demonstrated through statistical validation of data quality and through its pivotal role as the core dataset for the VK RecSys Challenge 2025. The work lowers barriers to realistic benchmarking and supports development of next-generation, platform-aware recommender systems.

Abstract

Short-video recommendation presents unique challenges, such as modeling rapid user interest shifts from implicit feedback, but progress is constrained by a lack of large-scale open datasets that reflect real-world platform dynamics. To bridge this gap, we introduce the VK Large Short-Video Dataset (VK-LSVD), the largest publicly available industrial dataset of its kind. VK-LSVD offers an unprecedented scale of over 40 billion interactions from 10 million users and almost 20 million videos over six months, alongside rich features including content embeddings, diverse feedback signals, and contextual metadata. Our analysis supports the dataset's quality and diversity. The dataset's immediate impact is confirmed by its central role in the live VK RecSys Challenge 2025. VK-LSVD provides a vital, open dataset to use in building realistic benchmarks to accelerate research in sequential recommendation, cold-start scenarios, and next-generation recommender systems.

Figures (2)

• Figure 1: Distributions of Key User and Item Statistics.
• Figure 2: Cosine Similarity Analysis of ALS Latent Factors Across User and Content Attributes