Search papers, labs, and topics across Lattice.
This paper introduces UniFormer, a unified model-centric scaling framework for industrial recommender systems that enhances efficiency by decomposing the modeling space into feature and task components. By employing a semantic-based tokenization scheme for user-item decoupling and multi-sequence cross-attention to capture heterogeneous behavior patterns, UniFormer significantly improves user engagement metrics in real-world applications. The results from extensive online A/B testing demonstrate notable increases in App Stay Time and Watch Time, underscoring the framework's effectiveness in practical settings.
UniFormer achieves up to a 1.113% increase in Watch Time, showcasing a breakthrough in user engagement for industrial recommendation systems.
Recently, substantial progress has been made in industrial recommendation through component-centric model scaling, where individual components such as behavior modeling, feature interaction, or task modeling are independently scaled to improve model capacity. Although recent methods such as HyFormer and OneTrans further explore cross-module co-scaling by jointly modeling behavior and interaction, their designs are still confined to the feature space and lack a unified model-centric scaling framework over the overall modeling space. In this paper, we propose UniFormer, an efficient and unified model-centric scaling framework for industrial recommender systems. To improve efficiency, UniFormer decomposes the overall modeling space into feature and task spaces, which are modeled by stacked Feature-space Interaction Modules and Task-space Interaction Modules, respectively. Moreover, UniFormer introduces semantic-based tokenization scheme to enable user-item decoupling, thereby achieving request-level inference acceleration. To prevent preference collapse, UniFormer employs multi-sequence cross-attention to separately capture heterogeneous behavior patterns, followed by the self-attention to enhance interaction modeling. Besides, dedicated multi-view FFNs are introduced to support flexible and scalable parameter scaling across different modeling components. Extensive online A/B testing in two production scenarios, Kuaishou and Kuaishou Lite, shows that UniFormer consistently improves user engagement and interaction metrics, achieving gains of +0.101%/+0.260% in App Stay Time and +0.729%/+1.113% in Watch Time, respectively.
