MuSeR improves long-sequence user recommendations with multi-interest modeling
MuSeR: Scalable Long-sequence Recommendation with Multi-interest Modeling
Information Retrieval
Summary
Recommender systems usually only look at a small part of a user's history, missing out on long-term interests. The authors present MuSeR, which compresses very long user histories and captures multiple interests by using both item IDs and text summaries from a language model. This system can handle millions of user actions efficiently and improves recommendations in real-world tests. It also reduces the time and cost to serve recommendations in apps like Baidu.
What this means in practice
- •For mobile app developers: Deploy MuSeR to enhance recommendation quality and efficiency in apps with large-scale user behavior histories.
- •For video streaming platform teams: Use MuSeR’s multi-interest extraction and multimodal alignment to better recommend diverse content types like short videos.
Authors
Yongkang Fu, Beining Bao, Yu Jiang, Xiangyu Zhao, Hongyang Wei, Guangxing Chen, Zuodong Yang, Shantao Li, Zonggang Wu, Yuqi Lu, Shouke Qin, Hanmeng Liu, Maolin Wang
Abstract
Ultra-long user behavior sequences carry rich signals of stable and diverse preferences, yet industrial recommender systems typically truncate histories to a few hundred actions under strict latency and memory budgets, leaving long-term interests under-utilized. Users also pursue multiple heterogeneous intents across modalities such as news, Q&A, and short video, which sparse ID embeddings alone struggle to represent. We present Multi-interest Sequence Representation (MuSeR), a retrieval framework built on the deployed MGS system, which integrates three components: (i) hierarchical temporal compression, which retains recent actions at full resolution while progressively pooling older segments, so that per-user histories of $10^{4}$-$10^{5}$ interactions fit within a fixed serving budget; (ii) disentangled multi-query interest extraction with orthogonality regularization; and (iii) multimodal semantic alignment, which augments sparse item IDs with textual summaries distilled from a large language model. For industrial deployment, MuSeR further adopts asynchronous user-representation refresh with adaptive caching and hierarchical beam-search retrieval across heterogeneous hardware. On three public benchmarks and a large-scale industrial dataset, MuSeR consistently improves Recall@$K$ over strong long-sequence and multi-interest baselines. In online A/B tests on Baidu APP's homepage feed, discovery feed, and short-video scenarios, MuSeR yields +0.26% daily active users and +0.89% total session duration (both statistically significant, p<0.05), alongside reduced serving latency and cost. Rather than proposing a new modeling primitive, our contribution is a system-level integration that makes long-term, multi-interest, and multimodal modeling jointly deployable in a real-time production pipeline, together with the engineering practices required to sustain it.