X-Rec Technical Report

2026-09-24

AI Summary

This study addresses the limitations of monolithic user-item representations in recommender systems, alongside the quantization errors and low-throughput bottlenecks inherent to autoregressive sequence ID models. To overcome these challenges, we propose a continuous-space recommendation distribution learning framework based on flow matching. Specifically, our method introduces anchor-conditioned decomposition, Riemannian flow matching for geometric alignment, and a late-interaction diffusion Transformer architecture to directly generate embeddings that trigger approximate nearest neighbor retrieval. Extensive evaluations demonstrate that the proposed approach achieves a 3.46× throughput improvement and significantly optimizes user interaction rates within TikTok’s vertical scenarios. The framework has been successfully deployed at industrial scale, validating its practical efficacy for large-scale recommendation tasks.

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%).

Problem

• recommender systems
• retrieval
• user-to-item
• semantic-ID autoregressive
• quantization error

Innovation

• Flow Matching
• Riemannian Flow Matching
• Anchor Conditioning
• Late-interaction Diffusion Transformer
• Continuous Embedding Space