This source only provides a headline — there is no body to show. Use “Read the original” below to read it at the source.
面向高效长上下文检索的加权状态汤化方法
AI summary
本文已被 NeurIPS 2026 的个性化、对齐、长期记忆人工智能系统研讨会(PALM)接收。状态空间模型(SSM)兼具快速推理、固定状态大小和有竞争力的语言建模性能。最近有人提出状态汤化(state souping)流程:SSM 独立编码文档、工具定义或记忆记录等 token 块,所得状态编码可缓存并重新组合,从而低成本地为下游查询构建上下文。然而,状态汤化和原始 SSM 在长且拥挤的……上都会退化。
Published by Apple Machine Learning
This site only summarises and aggregates; copyright stays with the original author. Headlines and summaries follow the source — where they differ, the original wins.