LongRetriever: Towards Ultra-Long Sequence based Candidate Retrieval for Recommendation

TL;DR

LongRetriever introduces a novel framework for incorporating ultra-long user sequences into the candidate retrieval stage of recommender systems, significantly improving performance and deployed at scale.

Contribution

The paper proposes in-context training and multi-context retrieval methods to effectively utilize ultra-long sequences during candidate retrieval, a previously under-explored area.

Findings

Significant online A/B testing improvements

Full deployment impacting billions of users

Effective candidate-specific interaction modeling

Abstract

Precisely modeling user ultra-long sequences is critical for industrial recommender systems. Current approaches predominantly focus on leveraging ultra-long sequences in the ranking stage, whereas research for the candidate retrieval stage remains under-explored. This paper presents LongRetriever, a practical framework for incorporating ultra-long sequences into the retrieval stage of recommenders. Specifically, we propose in-context training and multi-context retrieval, which enable candidate-specific interaction between user sequence and candidate item, and ensure training-serving consistency under the search-based paradigm. Extensive online A/B testing conducted on a large-scale e-commerce platform demonstrates statistically significant improvements, confirming the framework's effectiveness. Currently, LongRetriever has been fully deployed in the platform, impacting billions of users.