提出轻量时序层,区分信息新鲜度与主题演化,提升RAG检索准确率。
Freshness and the Limits of Heuristic Trend Detection in Temporal RAG
- 引入半衰期时效先验,优先召回最新相关数据
- 在硬测试集上最新项命中率达0.60,远超基线0.20
- 首次清晰拆分新鲜度与主题演化问题,适合安全领域研究者
我们提出一种轻量、模型无关的时序层用于RAG,并利用网络安全数据分离了通常被混淆的两个问题。针对新鲜度,半衰期时效先验在新数据中显著提升相关性,而仅用余弦相似度的基线得分仅为0.00;在困难的NVD CVE测试中,当最鲜资讯并非最相似时,该方法达到Latest@10为0.60,优于语义后排序的基线0.20,但仍存在部分性能受限且对参数敏感的问题。针对主题演化,启发式追踪器的宏平均F1仅为0.08,其根源在于标注规则而非聚类算法(HDBSCAN:0.10);修正标注规则后达0.49,去除聚类噪声后更达0.96。我们贡献了一个可复现的解耦框架,具有真实数据范围和参考实现。
原文摘要 · Abstract (English)
We present a lightweight, model-agnostic temporal layer for RAG and use cybersecurity data to separate two problems that are usually conflated. For freshness, a half-life recency prior surfaces the newest relevant item where a cosine-only baseline scores 0.00; on a hard NVD CVE test, where the freshest item is not the most similar, it reaches Latest@10 of 0.60 versus 0.20 for a semantic-then-newest baseline, but stays partial and parameter-sensitive. For topic evolution, a heuristic tracker's low 0.08 macro-F1 is driven by the labeling rule, not the clusterer (HDBSCAN: 0.10; fixing the rule alone reaches 0.49, and 0.96 without clustering noise). We contribute a reproducible decoupling of the two, with honest real-data scope and a reference implementation.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品;中文卡片由大模型生成,请以原文为准。