When Classic Cache Policies Fail: Learning-Augmented Replacement for Semantic Retrieval Buffers

  • 类型:arxiv
  • 标识:2607.00394
  • 链接:https://arxiv.org/abs/2607.00394
  • 主分类:rag
  • 形态:method
  • 被引:1
  • 被引来源:Semantic Scholar
  • S2被引:1
  • OpenAlex被引:0
  • 影响力被引:1
  • TLDR:SOLAR is proposed, a learning-augmented framework that derives modification timing from regret accumulation and content selection from Bayesian online learning over implicit retrieval feedback and achieves a constant competitive ratio, independent of cache size and horizon.
  • OpenAlex ID:W7167023563
  • OpenAlex DOI:10.48550/arxiv.2607.00394
  • DOI:10.48550/arxiv.2607.00394
  • DOI来源:OpenAlex
  • 开放获取:green
  • 开放获取链接:https://doi.org/10.48550/arxiv.2607.00394
  • OpenAlex更新:2026-07-19
  • 待LLM分类:否
  • 标题中文:当经典缓存策略失效时:面向语义检索缓冲区的学习增强替换
  • TLDR中文:本文提出 SOLAR,一种学习增强框架,从 regret 累积中推导修改时机,并基于隐式检索反馈的贝叶斯在线学习进行内容选择,实现与缓存大小和时域无关的常数竞争比。
  • 来源文件
  • /inbox/tom/_candidates/2026-07-08-agent-rag-longcontext-candidates.json
  • [S2 enrich]
  • [OpenAlex backfill]