Newshot

斯坦福开源 Memorilla:面向冻结 LLM 的潜在记忆模块

AI summary

斯坦福研究人员开源了面向 LLM 的记忆系统 Memorilla,它能读取整个文档集合并只向冻结的 LLM 提供 16 个记忆 token。在 FactKG 上,冻结的 Qwen3-8B 搭配 Memorilla 得分为 84.33%,而将所有文档放入上下文时为 80.43%,且文档 token 减少了 108 倍。代码、训练配方和检查点以 Apache 2.0 协议公开。

This source only provides a headline — there is no body to show. Use “Read the original” below to read it at the source.

Published by MarkTechPost
This site only summarises and aggregates; copyright stays with the original author. Headlines and summaries follow the source — where they differ, the original wins.
Read the original ↗