论文73°

Meta AI 用记忆智能体缓解长上下文遗忘,Claude Sonnet 4.5 成绩升至 45.9%

精选理由

Meta 给智能体配了个记忆员,专记它犯过的错并适时提醒,把 Claude Sonnet 4.5 成绩从 37.6% 拉到 45.9%,做智能体的可以看看。

Meta AI 发现长上下文会让智能体遗忘早期犯过的错误,即 context rot 问题。解决方案是给主行动智能体配一个专职记忆智能体,由后者记录工具使用情况与过往错误的结构化笔记。记忆智能体只在关键时机注入简短提醒,而不是塞满整个上下文。该方法把 Claude Sonnet 4.5 的基准成绩从 37.6% 提升到 45.9%。

原文 · DeepLearning.AI

Long contexts cause AI agents to forget early mistakes. Meta AI paired primary action agents with dedicated memory agents to fix context rot: 📝 Keeps structured notes on tools and past errors 🎯 Injects short reminders only at crucial moments 📈 Raised Claude Sonnet 4.5 benchmarks from 37.6% to 45.9% Read the full technical analysis hubs.la/Q04yf27q0 4F #DeepLearningAI g #AIAgents n #LLMs LMs 💬 5 🔄 1 ❤️ 21 👀 1470 📊 8 ⚡