跳到正文
arXiv cs.AI· Minsang Kim, Junpyo Lim, Heeje Cho, Seungjun Baek·· 2 小时前AI 评分35

StitchRAG:基于扩散拼接为 RAG 构建紧凑且保留证据的上下文

StitchRAG: Constructing Compact and Evidence-Preserving Contexts for RAG with Diffusion-Based Stitching

AI 导读

arXiv 论文提出 StitchRAG,一个为 RAG 构建紧凑上下文的框架,在动态压缩检索段落的同时保留忠实、信息密集的内容。该方法按段落自适应压缩比例,提取答案候选与查询相关关键词作为锚点,并用文本扩散将句子拼接成连贯上下文。在五个开放域 QA 基准上,StitchRAG 持续优于多种多步与压缩基线方法。

正文 · 原文

View PDF HTML (experimental)

Abstract:Retrieval-augmented generation (RAG) improves factuality by grounding large language models (LLMs) in external evidence. However, long retrieved contexts incur computational overhead and can even hurt performance due to noisy or irrelevant content. We propose \textsc{StitchRAG}, a framework for constructing compact context in RAG. \textsc{StitchRAG} builds a faithful, information-dense context while dynamically compressing the retrieved passages. We adapt the size reduction ratio to each passage by formulating a utility maximization trading-off the retained information and computation costs. We then extract answer candidates and query-related keywords from passages as anchors and collect passage sentences containing anchors under the derived size constraint. Finally, we connect or stitch the collected sentences together using text diffusion to form a coherent and natural context. Experiments on five open-domain QA benchmarks show that \textsc{StitchRAG} consistently outperforms diverse multi-step and compression baselines. Extensive ablations show the effectiveness of dynamic size allocation, anchor preservation, and diffusion-based stitching.
Subjects: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
Cite as: arXiv:2611.22766 [cs.CL]
  (or arXiv:2611.22766v1 [cs.CL] for this version)
  https://doi.org/10.48550/arXiv.2611.22766

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Minsang Kim [view email]
[v1] Tue, 17 Nov 2026 13:33:26 UTC (4,750 KB)

来源:arXiv cs.AI · arxiv.org