Method
SASRec
Self-attentive sequential recommender over user interaction history.
7 cases from 6 companies apply SASRec. 4 were checked in an online A/B test; 1 report results in production.
Typical tasks: Recommend from what the user just did, Retrieve and rank from millions of items, Balance relevance and diversity, Reduce ML cost and 1 more. Companies: Yandex, Skoltech, Avito, VK, T-Bank and 1 more. Often combined with: Sampled softmax, Two-tower model, Gradient boosting, ALS, Collaborative filtering and 16 more.
Scaling Recommender Systems![]() | ![]() | media-streaming | 2025-05 |
Scaling Loss Functions for Sequential Recommendations with Large Item Catalogs![]() | ![]() | general | 2025-05 |
Speeding Up Inference on Graph Models for Warm Start (AIRI)![]() | ![]() | general | 2025-05 |
Cut Cross Entropy: Efficient Training of Transformers for Recommender Systems![]() | ![]() | general | 2025-05 |
| How We Moved from a Markov Chain to a SASRec Model | classifieds | 2024-05 | |
| Transforming Recommendations | social | 2024-05 | |
Reading Between the Lines: How to Use ML to Curate Personalized Collections Better Than Editors![]() | ![]() | fintech | 2024-05 |



