斯坦福发了个AutoMem，给智能体装上可训练的记忆模块，让32B模型打平Claude Opus 4.5。省去调任务动作，光改记忆就提升2-4倍，搞长程任务的别错过。
斯坦福大学提出AutoMem框架，将智能体记忆管理从固定模块转变为可训练技能。模型自主决定编码、检索与笔记组织，文件系统操作作为一等动作。AutoMem通过两循环自动化：强LLM重写记忆结构，智能体历史记忆决策作为训练信号。仅优化记忆使基础智能体在Crafter、MiniHack、NetHack上提升2-4倍。32B开源模型因此与Claude Opus 4.5、Gemini 3.1 Pro Thinking性能相当。
// AutoMem // I quite like this idea of metamemory. (bookmark it) This new research from Stanford...
// AutoMem // I quite like this idea of metamemory. (bookmark it) This new research from Stanford treats agent's memory management as a trainable skill instead of a fixed module. The model decides what to encode, when to retrieve, and how to organize its own notes, with file-system operations promoted to first-class actions right alongside task actions. AutoMem automates this on two loops. A strong LLM reviews full trajectories and rewrites the memory structure (prompts, schemas, action vocabulary). Then the agent's own good memory decisions across episodes become training signal to sharpen its proficiency. Optimizing memory alone, without touching task-action behavior, lifts the base agent 2x to 4x on Crafter, MiniHack, and NetHack. That is enough to make a 32B open model competitive with Claude Opus 4.5 and Gemini 3.1 Pro Thinking. For long-horizon agents, memory is a high-leverage objective you can train for on its own. Paper: arxiv.org/abs/2607.01224 Learn to build effective AI agents in our academy: academy.dair.ai 💬 7 🔄 11 ❤️ 51 👀 5245 📊 25 ⚡
- arXiv cs.AI07-01 17:57原文