X-Rec improves recommendation speed and accuracy with continuous embedding generation
X-Rec Technical Report
Information Retrieval
Summary
Recommender systems suggest items like videos or products by predicting what a user might want next. Traditional methods either simplify user interests into a few fixed signals or use complex sequences that are slow to process. The authors propose X-Rec, which generates item features in a continuous space to better capture varied user preferences and speed up predictions. They tested X-Rec on real-world data and found it matches the best accuracy methods while being over three times faster. This approach has been successfully applied at TikTok, improving user engagement on specific content types.
What this means in practice
- •For recommendation system engineers: Enhance streaming content recommendation by generating continuous embeddings for faster and more accurate item retrieval.
- •For content platform developers: Deploy improved retrieval methods to increase vertical and general user engagement in video recommendation environments.$Commercial implications: Enables a scalable retrieval component that directly improves engagement metrics on platforms like TikTok, potentially increasing revenue.
Authors
Chenglei Shen, Chenzhe Huang, Dong Jiang, Hongjie Gao, Jue Zhang, Kun Xú, Lincan Cai, Nan Zhuang, Pan Zhang, Shi Chen, Shunchi Zhang, Xiaoyu Ye, Yang Jin, Yu Zhang, Zhenwei An, Zhongtao Jiang, Zhiwei Wang, Kun Xǔ
Abstract
Recent advances in generative modeling have reshaped recommender systems by formulating recommendation as a next-item generation problem. Existing retrieval approaches primarily follow two paradigms: user-to-item (U2I) methods represent user context using one or a few deterministic embeddings, which limits the ability to capture diverse and multi-mode interests, while semantic-ID-based autoregressive (SID-AR) methods model more expressive distributions but suffer from quantization errors and the low throughput of sequential decoding. To address these limitations, we propose X-Rec to directly learn the recommendation distribution in the continuous item embedding space through flow matching and generate embedding triggers for approximate nearest neighbor retrieval. X-Rec incorporates three key designs to make this formulation effective and efficient. First, we introduce anchor conditioning to decompose generation into coarse semantic-region selection and fine-grained refinement. Second, we adopt Riemannian flow matching to align generative trajectories with the hyperspherical geometry of item embeddings. Third, we design a late-interaction diffusion Transformer that restricts repeated velocity-field estimation to the final Transformer layer. On a streaming benchmark, X-Rec substantially outperforms U2I baselines, matches the retrieval quality of SID-AR methods, and delivers 3.46x higher inference throughput than SID-AR. X-Rec has also been deployed as a new retrieval source for a specific vertical content on TikTok, where two consecutive launches have yielded significant improvements in both vertical engagement (+4.1484%) and general engagement (+0.0111%).