让大模型主动学习并管理记忆，提升长期推理能力。
Towards Autonomous Memory Agents
- 用低成本多阶段方法自动获取、验证和整理知识。
- 在多个基准上超越已有方法，最高提升14.6分。
- 适合需要持续学习与自我修正的智能体应用。
近期的记忆代理通过将经验与对话历史提取至外部存储，提升了大语言模型的表现，实现了低开销的上下文组装和在线记忆更新，无需昂贵的模型训练。然而，现有方案仍被动响应；记忆增长受限于偶然可用的信息，且在不确定时极少主动寻求外部输入。本文提出自主记忆代理，以最低成本主动获取、验证和维护知识。U-Mem 通过（i）成本感知的知识提取级联，从廉价的自生成/教师信号逐步升级到工具验证的研究内容，仅在必要时引入专家反馈；（ii）语义感知的汤普森采样，在记忆探索与利用之间取得平衡，并缓解冷启动偏差。在可验证与不可验证的基准测试中，U-Mem 均显著优于先前记忆基线，甚至超过基于强化学习的优化方法：在 HotpotQA（Qwen2.5-7B）上提升14.6点，在 AIME25（Gemini-2.5-flash）上提升7.33点。
原文摘要 · Abstract (English)
Recent memory agents improve LLMs by extracting experiences and conversation history into an external storage. This enables low-overhead context assembly and online memory update without expensive LLM training. However, existing solutions remain passive and reactive; memory growth is bounded by information that happens to be available, while memory agents seldom seek external inputs in uncertainties. We propose autonomous memory agents that actively acquire, validate, and curate knowledge at a minimum cost. U-Mem materializes this idea via (i) a cost-aware knowledge-extraction cascade that escalates from cheap self/teacher signals to tool-verified research and, only when needed, expert feedback, and (ii) semantic-aware Thompson sampling to balance exploration and exploitation over memories and mitigate cold-start bias. On both verifiable and non-verifiable benchmarks, U-Mem consistently beats prior memory baselines and can surpass RL-based optimization, improving HotpotQA (Qwen2.5-7B) by 14.6 points and AIME25 (Gemini-2.5-flash) by 7.33 points.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。