CoinRAG：通过上下文信息片段 KV 缓存复用优化长上下文 RAG

AI 摘要

CoinRAG 是一种针对长上下文检索增强生成（RAG）的优化方法，通过复用细粒度的语义片段缓存而非完整块，在低预填充延迟约束下提升效率与准确性。在 LongBench 多跳问答任务上，CoinRAG 平均相对提升 F1 分数 5.3%，并实现了新的帕累托前沿。

正文节选

CoinRAG: Contextualized Information Nugget KV Cache Reuse for Long-Context RAG Abstract CoinRAG improves retrieval-augmented generation efficiency and accuracy by reusing fine-grained semantic nugget caches instead of full chunks. Recent optimization studies on Retrieval-Augmented Generation (RAG) have exploited chunk-level KV cache reuse to avoid processing long retrieved contexts for higher efficiency, while significant information redundancy and noise still remain in the coarse-grained chunks

发布时间：—

抓取时间：2026-08-13 08:43

来源机构：Hugging Face