VK Research: classical recommendation algorithms can account for users' future interests
VK
AI VK Research presented a new approach that lets classical, non-neural recommendation algorithms factor in how a current recommendation will influence a user's future preferences, not just present interests. The method was accepted at the Customer Journey workshop of KDD 2026 and showed improved accuracy on open datasets including VK-LSVD.
Why it matters
The approach can be integrated into existing production recommender systems without a full architectural rebuild, and VK says it is already deployed for several applied tasks in its ecosystem, illustrating a low-overhead route to planning-style objectives in industrial recommendation pipelines.
Importance: 2/5
Notable applied research from a major Russian tech company, confirmed by official release and independent media.