超越最终嵌入：从Token级证据中检索记忆
原标题：Beyond the Final Embedding: Retrieving Memories from Token-Level Evidence
AI 摘要
该论文提出一种无需额外训练的检索方法，利用冻结嵌入模型已计算的上下文 token 表示进行稀疏 token 匹配，再与整体记忆相似度结合来排序记忆。在八个检索基准和多个模型家族上均取得提升，在 Qwen3-Embedding-4B 上配合查询适配器可将 Recall@10 和 NDCG@10 平均提升 13.3 和 14.3 个百分点，并转化为长期记忆问答准确率提升。
正文节选
Beyond the Final Embedding: Retrieving Memories from Token-Level Evidence Abstract Memory-augmented assistants retrieve evidence from stored conversations and documents to answer queries. Most embedding-based retrievers compare single-vector representations of queries and memories, even when relevance depends on a small part of the content. Such local evidence can be matched to queries using the contextual token representations already computed by the same embedding models. Our method uses spars