MixFormer: Co-Scaling Up Dense and Sequence in Industrial Recommenders

TL;DR

MixFormer is a unified Transformer architecture for industrial recommender systems that jointly models sequence behaviors and feature interactions, enabling better co-scaling, expressiveness, and efficiency.

Contribution

The paper introduces MixFormer, a novel unified Transformer model that co-scales dense features and sequences, addressing the limitations of decoupled designs in recommendation systems.

Findings

Outperforms existing models in accuracy on large-scale datasets.

Reduces inference latency through user-item decoupling strategies.

Achieves significant improvements in user engagement metrics in online tests.

Abstract

As industrial recommender systems enter a scaling-driven regime, Transformer architectures have become increasingly attractive for scaling models towards larger capacity and longer sequence. However, existing Transformer-based recommendation models remain structurally fragmented, where sequence modeling and feature interaction are implemented as separate modules with independent parameterization. Such designs introduce a fundamental co-scaling challenge, as model capacity must be suboptimally allocated between dense feature interaction and sequence modeling under a limited computational budget. In this work, we propose MixFormer, a unified Transformer-style architecture tailored for recommender systems, which jointly models sequential behaviors and feature interactions within a single backbone. Through a unified parameterization, MixFormer enables effective co-scaling across both dense capacity and sequence length, mitigating the trade-off observed in decoupled designs. Moreover, the integrated architecture facilitates deep interaction between sequential and non-sequential representations, allowing high-order feature semantics to directly inform sequence aggregation and enhancing overall expressiveness. To ensure industrial practicality, we further introduce a user-item decoupling strategy for efficiency optimizations that significantly reduce redundant computation and inference latency. Extensive experiments on large-scale industrial datasets demonstrate that MixFormer consistently exhibits superior accuracy and efficiency. Furthermore, large-scale online A/B tests on two production recommender systems, Douyin and Douyin Lite, show consistent improvements in user engagement metrics, including active days and in-app usage duration.