Embedding Items at Scale: Comparing GNN-Based and ID-Based Item Embeddings in the Yandex Ecosystem

2026-07-28

AI Summary

This study addresses the lack of systematic comparison between pre-trained graph neural network (GNN) embeddings and end-to-end trainable ID embeddings in large-scale industrial recommender systems, particularly regarding their trade-offs in cost and effectiveness. For the first time, the authors conduct a comprehensive evaluation of these two item embedding strategies within a Transformer-based sequential recommendation framework across three real-world production environments at Yandex: Market, Music, and Lavka. Experimental results demonstrate that under data-rich conditions, GNN-pretrained embeddings do not significantly outperform ID embeddings; however, in low-resource settings, they yield notable performance gains. To foster reproducible research, the authors publicly release the associated datasets and codebase.

Abstract

Transformer-based sequential recommendation models, which process sequences of user-item interactions, rely heavily on the item embedding strategy. Existing approaches either use pretrained item embeddings or learn them end-to-end with the transformer. To the best of our knowledge, no prior work has compared these options from both cost and quality perspectives in a large-scale industrial setting. This paper is a case study that compares pretrained industrial graph neural network item embeddings with end-to-end trainable item embeddings across two mature production recommendation systems at Yandex: Yandex Market and Yandex Music. We additionally evaluate both approaches on a low-resource dataset sampled from Yandex Lavka production logs, for which both the data and code are publicly available for demonstration purposes. Our results show that a separate pretraining stage helps when training data is limited, but provides no worthwhile benefit for large-scale models trained on extensive datasets.

Problem

• item embeddings
• graph neural networks
• sequential recommendation
• pretraining
• large-scale recommendation

Innovation

• item embeddings
• graph neural networks
• sequential recommendation
• pretraining
• industrial-scale evaluation