用外部记忆库让大模型不改参数就能持续学习，效率提升2800倍。
Learning Hierarchical Procedural Memory for LLM Agents through Bayesian Selection and Contrastive Refinement
- 用分层程序记忆存储可复用技能，不修改大模型参数
- 在4个基准上平均准确率78.1%，未见任务达90.3%
- 构建记忆仅需56秒，压缩2851条轨迹为187个程序
我们提出MACLA框架，通过外部分层程序记忆实现推理与学习解耦。该框架在冻结大语言模型的前提下，从轨迹中提取可复用程序，利用贝叶斯后验追踪可靠性，通过期望效用评分选择动作，并基于成功与失败对比进行程序优化。在ALFWorld、WebShop、TravelPlanner、InterCodeSQL四个基准测试中，MACLA平均性能达78.1%，在ALFWorld未见任务上达到90.3%准确率，正向泛化率3.1%。系统在56秒内完成记忆构建，比当前最优的参数微调基线快2800倍，将2851条轨迹压缩为187个程序。实验表明，结合贝叶斯选择与对比优化的结构化外部记忆，可在不更新模型参数的情况下实现高效、可解释且持续进化的智能体。
原文摘要 · Abstract (English)
We present MACLA, a framework that decouples reasoning from learning by maintaining a frozen large language model while performing all adaptation in an external hierarchical procedural memory. MACLA extracts reusable procedures from trajectories, tracks reliability via Bayesian posteriors, selects actions through expected-utility scoring, and refines procedures by contrasting successes and failures. Across four benchmarks (ALFWorld, WebShop, TravelPlanner, InterCodeSQL), MACLA achieves 78.1 percent average performance, outperforming all baselines. On ALFWorld unseen tasks, MACLA reaches 90.3 percent with 3.1 percent positive generalization. The system constructs memory in 56 seconds, 2800 times faster than the state-of-the-art LLM parameter-training baseline, compressing 2851 trajectories into 187 procedures. Experimental results demonstrate that structured external memory with Bayesian selection and contrastive refinement enables sample-efficient, interpretable, and continually improving agents without LLM parameter updates.
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。