不浪费：通过多头解码以结构化人类先验引导生成式推荐系统
Don't Waste It: Guiding Generative Recommenders with Structured Human Priors via Multi-head Decoding
November 13, 2025
作者: Yunkai Zhang, Qiang Zhang, Feng, Lin, Ruizhong Qiu, Hanchao Yu, Jason Liu, Yinglong Xia, Zhuoran Yu, Zeyu Zheng, Diji Yang
cs.AI
摘要
为提升推荐系统的长期用户满意度，优化除准确性外的多样性、新颖性及个性化等目标至关重要。工业界从业者已积累了大量结构化领域知识（如物品分类体系、时序模式），我们将其定义为人因先验。这类知识通常通过排名或后排名阶段的后期调整来实现，但该方法始终与核心模型学习相割裂——这在行业向端到端生成式推荐基础模型转型的背景下尤显不足。另一方面，许多针对超准确性目标的方法常需针对特定架构进行修改，并以完全无监督的方式学习用户意图，从而舍弃了这些宝贵的人因先验。
为有效利用经年积累的人因先验，我们提出了一种与主干模型无关的框架，将其无缝集成到生成式推荐器的端到端训练中。借鉴高效大语言模型解码策略，我们设计了轻量级的先验条件适配头，引导模型沿人类可理解的维度（如交互类型、长短期兴趣）解构用户意图。同时引入分层组合策略，以建模不同先验类型间的复杂交互。在三个大规模数据集上的实验表明，我们的方法显著提升了准确性及超准确性目标。研究还揭示，人因先验能使主干模型更有效地利用长上下文窗口与大模型规模。
（注：译文采用学术论文常见的被动语态与名词化表达，通过"将其定义为人因先验"等句式实现术语准确对应；使用"尤显不足""经年积累"等四字格提升文本凝练度；通过分号与破折号构建复杂逻辑关系，符合中文科技论文表达规范。）
English
Optimizing recommender systems for objectives beyond accuracy, such as diversity, novelty, and personalization, is crucial for long-term user satisfaction. To this end, industrial practitioners have accumulated vast amounts of structured domain knowledge, which we term human priors (e.g., item taxonomies, temporal patterns). This knowledge is typically applied through post-hoc adjustments during ranking or post-ranking. However, this approach remains decoupled from the core model learning, which is particularly undesirable as the industry shifts to end-to-end generative recommendation foundation models. On the other hand, many methods targeting these beyond-accuracy objectives often require architecture-specific modifications and discard these valuable human priors by learning user intent in a fully unsupervised manner.
Instead of discarding the human priors accumulated over years of practice, we introduce a backbone-agnostic framework that seamlessly integrates these human priors directly into the end-to-end training of generative recommenders. With lightweight, prior-conditioned adapter heads inspired by efficient LLM decoding strategies, our approach guides the model to disentangle user intent along human-understandable axes (e.g., interaction types, long- vs. short-term interests). We also introduce a hierarchical composition strategy for modeling complex interactions across different prior types. Extensive experiments on three large-scale datasets demonstrate that our method significantly enhances both accuracy and beyond-accuracy objectives. We also show that human priors allow the backbone model to more effectively leverage longer context lengths and larger model sizes.