Tag
This Stanford research paper introduces AutoMem, a framework that treats agent memory management as a trainable skill. By optimizing memory structure and proficiency separately, AutoMem improves base agent performance 2x-4x on long-horizon tasks, enabling a 32B open-weight model to compete with frontier systems like Claude Opus 4.5 and Gemini 3.1 Pro Thinking.
AutoMem introduces a framework that automates learning of memory management as a trainable skill for LLMs, improving performance on long-horizon tasks by 2x-4x through optimizing memory structure and proficiency.