基本信息
- 来源: arxiv
- 原始来源: http://arxiv.org/abs/2608.07458v1
- 发布域名: arxiv.org
- 分类: cs.CL
- 作者: Gyuwan Kim、Cheoneum Park、Tao Yang
要点解读
这是什么
该方法通过在检索得到的段落中定位与查询相关的细粒度语义单元，并把它们对应的 KV 表示与段落级上下文组合，实现细粒度缓存的复用，从而在保持较低预填充延迟的同时提升答案质量。
用在哪里
适用于需要处理长上下文的检索增强生成系统，特别是多跳问答类任务。研发长上下文模型或优化检索流水线的人员可能会关注此类技术，以在效率和准确性之间取得更好的平衡。
可以推断的
推测：在实际部署中，缓存的细粒度划分可以降低对完整块重新编码的计算开销，从而减少响应延迟。
推测：该技术的实现需要额外的检索阶段来筛选相关语义单元，可能增加系统设计的复杂度。
来源摘要/节选
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. This paper optimizes the Pareto frontier under low prefill latency constraints while maximizing accuracy by proposing CoinRAG (Contextualized Information Nugget KV Cache Reuse for Long-Context RAG). The name metaphorically reflects our core mechanism: much like assembling small tokens (or “coins”) to accumulate a larger value, CoinRAG compositionally reuses offline-computed, fine-grained nugget caches to form a learned contextual representation efficiently in a more semantically relevant but compact manner. Specifically, instead of full-chunk encoding, CoinRAG identifies query-relevant semantic units within retrieved chunks through two-stage retrieval and seamlessly assembles their sliced KV representations with a chunk-level context. Extensive evaluations on LongBench multi-hop question answering tasks demonstrate that CoinRAG significantly reduces operational costs and outperforms the other baselines with a new Pareto frontier and an average 5.3% relative improvement in answer quality (F1) under a standard fast prefill latency budget.
来源说明
当前保存的是来源摘要，不代表论文全文。请以原始来源为准。
「要点解读」由 AI Stack 依据上方已保存内容整理，不代表来源的完整表述；标注「推测：」的判断来自编辑，不是来源陈述。