用少量可学习的隐式枢纽，高效建模推荐系统中异构特征的复杂交互。
HubMixer: Progressive Latent Hub Mixing for Parameter-Efficient Feature Interaction in Recommendation
- 引入可学习的隐式枢纽，分三步组织特征交互：归纳、交互、读出。
- 在快手短视频招聘业务中提升简历提交转化率5.48%，显著优于现有模型。
- 适合追求高效率与强表达力的工业级推荐系统，尤其擅长处理异构特征。
有效学习特征交互是工业推荐与广告排序系统的核心。近期的 token 混合架构通过轻量级混合算子简化自注意力机制，提升了硬件效率并支持大规模部署。然而，推荐 token 具有根本性异质性：用户画像、物品属性、行为序列、上下文特征、统计信号及业务侧特征存在于不同语义空间，且以稀疏、样本特定的方式交互。直接在原始异质 token 空间中混合可能参数效率低下，因模型需隐式发现哪些特征组应交互以及如何路由。本文提出 HubMixer，一种用于推荐中特征交互的参数高效潜在枢纽混合架构。不同于直接混合原始特征 token，HubMixer 引入一组小规模可学习的潜在枢纽，通过‘归纳-交互-读出’范式组织特征交互。首先，枢纽归纳将异质 token 汇聚为紧凑的潜在枢纽，通过交叉注意力让枢纽查询输入 token；其次，在更清洁的潜在枢纽空间中执行高阶交互；最后，基于 token 条件的读出机制使每个原始 token 可选择性地从交互后的枢纽中读取信息，注入全局交互语义的同时保留各 token 的字段身份。在工业推荐任务上的大量离线实验表明，HubMixer 超越当前最优模型。在快手短视频招聘业务的在线 A/B 测试中，简历提交转化率提升 5.48%，已全面投入生产。
原文摘要 · Abstract (English)
Learning effective feature interactions is central to industrial recommendation and advertising ranking systems. Recent token-mixing architectures simplify self-attention with lightweight mixing operators, improving hardware efficiency and enabling large-scale deployment. However, recommendation tokens are fundamentally heterogeneous: user profiles, item attributes, behavioral sequences, context features, statistical signals, and business-side features live in different semantic spaces and interact in sparse, sample-specific patterns. Directly mixing all tokens in the raw heterogeneous token space may therefore be parameter-inefficient, as the model must implicitly discover which feature groups should interact and how such interactions should be routed. In the paper, we propose HubMixer, a parameter-efficient latent hub mixing architecture for feature interaction in recommendation. Instead of directly mixing raw feature tokens, HubMixer introduces a small set of learnable latent hubs to organize feature interactions through an `induction--interaction--readout` paradigm. First, hub induction summarizes heterogeneous tokens into compact latent hubs, where latent hubs query input tokens through cross-attention. Second, hub interaction performs high-order interaction in the cleaner latent hub space. Third, token-conditioned readout lets each original token selectively read from the interacted hubs, injecting global interaction semantics while preserving token-level field identity. Extensive offline experiments on industrial recommendation tasks show that HubMixer outperforms the SOTA models. Online A/B testing in the Kuaishou short-video recruitment business further shows a statistically significant 5.48% improvement in resume submission conversion rate, and HubMixer has been fully deployed in production.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。