生成即排序：基于统一语义-协同ID的端到端生成式 slate 推荐
Once Generated, Ranked: End-to-End Generative Slate Recommendation with Unified Semantic-Collaborative IDs
浏览论文内容
中文总结 AI 辅助
研究针对现有生成式 slate 推荐的缺陷，提出 OGR 框架，引入 TUSID 与 SPA 方法，在离线实验和快手在线测试中均取得显著性能提升。
中文摘要 AI 辅助
slate 推荐以 slate 而非单个物品作为推荐单元，需联合优化物品交互与 slate 效用。现有方法通常将候选生成与排序分离，且优化仅针对检索到的候选。带语义ID（SIDs）的生成式推荐为端到端推荐提供了路径，但现有 SID 构建常缺乏推荐感知语义与有效局部协同信号，且下一个 token 预测与 slate 级目标不匹配。我们提出 OGR，一个直接生成有序 slate 的端到端框架，即“Once Generated, Ranked”。OGR 首先引入 TUSID，其自适应融合物品特定语义与局部协同信息至分层 SIDs；接着采用列表式偏好规划与流水线式位置 SID 解码，以在生成有序 slate 时建模全局偏好与物品间依赖关系。我们进一步提出 SPA，一种奖励引导的保守策略优化方法，使生成的 slate 与用户偏好对齐，而非仅模仿似然。离线实验显示，OGR 优于代表性基线，在工业和公开数据集上分别取得 48.2% 和 27.2% 的相对 NDCG@5 提升；在快手开展的在线 A/B 测试进一步使有效观看量提升 1.120%。
英文摘要
Slate recommendation treats a slate rather than an individual item as the recommendation unit, requiring joint optimization of item interactions and slate utility. Existing approaches typically separate candidate generation from ranking and restrict optimization to retrieved candidates. Generative recommendation with Semantic IDs (SIDs) offers a path to end-to-end recommendation, but existing SID construction often lacks recommendation-aware semantics and effective local collaborative signals, while next-token prediction is misaligned with slate-level objectives. We propose OGR, an end-to-end framework that directly generates ordered slates-"Once Generated, Ranked." OGR first introduces TUSID, which adaptively fuses item-specific semantic and local collaborative information into hierarchical SIDs. It then uses list-wise preference planning and pipelined position-wise SID decoding to model global preferences and inter-item dependencies while generating ordered slates. We further propose SPA, a reward-guided conservative policy optimization method that aligns generated slates with user preferences beyond likelihood imitation. Offline experiments show that OGR outperforms representative baselines, with 48.2% and 27.2% relative NDCG@5 gains on industrial and public datasets, respectively. Online A/B testing on Kuaishou further yields a 1.120% improvement in Effective Views.
