LinkedIn 提出 ESP：动态多目标检索的嵌入子空间划分
原标题：Embedding Subspace Partitioning for Dynamic Multi-Objective Retrieval
AI 摘要
LinkedIn 提出 Embedding Subspace Partitioning (ESP) 检索框架，将嵌入分解为任务感知子空间，用可在线调节权重的子空间相似度加权和替代单一内积，从而在服务时动态调整多目标权衡而无需重训练。该方法在 MS MARCO 基准上覆盖更宽的 Pareto 前沿，并已在 LinkedIn 求职匹配平台部署两年，服务超 7000 万周活用户；7 天在线 A/B 测试中仅调整服务时权重即带来 +4.0% 收入、+3.95% 职位申请和 8.95% 不匹配率改善。
正文节选
Embedding Subspace Partitioning for Dynamic Multi-Objective Retrieval Abstract. Modern industrial recommender systems must optimize across competing objectives, balancing semantic relevance with business metrics such as engagement and revenue. While bi-encoders dominate large-scale retrieval due to their efficiency, they collapse these heterogeneous signals into a single static embedding space. This design creates a fundamental limitation: once trained, the retriever cannot adapt to shifting obj