MemRetriever：从长期记忆中学习搜索、反思与检索
原标题：MemRetriever: Learning to Search, Reflect, and Retrieve from Long-Term Memory
AI 摘要
MemTensor 提出 MemRetriever，一种将长期记忆检索建模为多步搜索过程的智能体式检索模型，通过并行搜索、串行搜索和反思去噪三种动作，主动规划检索与停止时机。研究使用 ReAct 风格轨迹做监督预热，并用 GRPO 强化学习优化证据覆盖、降噪和答案充分性。在 LOCOMO、LongMemEval、HotpotQA、MuSiQue 和 2WikiMultiHopQA 上，MemRetriever 优于静态 top-k 检索基线和仅监督变体，其中 MemRetriever-4B-RL 在 LongMemEval 主检索指标上超过 DeepSeek-v4-Flash。
正文节选
\ul 1]MemTensor (Shanghai) Technology *]Equal Contribution \titlefont MemRetriever: Learning to Search, Reflect, and Retrieve from Long-Term Memory Abstract Long-term memory is essential for personalized agents, but effective memory use depends not only on storing information, but also on retrieving the right evidence at the right time. Existing memory retrieval systems commonly rely on static top- retrieval, which issues a single query, returns a fixed number of memories, and directly passes th