txtai 推出 LEMUR 与均值中心化，提升晚期交互检索效率
原标题：LEMUR and Mean Centering for Late-Interaction Retrieval in txtai
AI 摘要
txtai 引入了 LEMUR（Learned Multi-Vector Retrieval）和可配置的均值中心化，以改进晚期交互检索。LEMUR 学习多向量表示的固定维度编码，使晚期交互模型能使用标准向量索引；均值中心化解决了 token 向量各向异性问题。在 BEIR 数据集上，LEMUR 在相同向量维度下相比 MUVERA 显著提升 NDCG@10，但效果依赖具体配置，且默认 IVF 索引会降低性能。
正文节选
Late-interaction models preserve a useful level of detail: instead of collapsing a query or document into one embedding immediately, they retain a vector for each token. A MaxSim score compares every query token with the document tokens, keeps the strongest match for each query token, and sums those matches. The tradeoff is operational. A multi-vector representation does not fit as naturally into the fixed-vector indexes used by a conventional dense retrieval pipeline. LEMUR closes that gap in t