ReTAG通过检索与主题增强的图结构，高效实现全文信息整合回答。
ReTAG: Retrieval-Enhanced, Topic-Augmented Graph-Based Global Sensemaking
- 构建主题相关子图并检索摘要，提升信息关联性
- 相比基线，响应质量更高且推理时间大幅降低
- 适合需要跨文档理解的复杂问答场景
近年来，问答任务在多跳推理方面取得显著进展。然而，全局理解——即从整个语料库中综合信息回答问题——仍面临重大挑战。先前基于图的方法缺乏检索机制、主题针对性，且推理成本高昂。为解决这些问题，我们提出ReTAG：一种检索增强、主题增强的图框架，通过构建主题特定子图并检索相关摘要来生成答案。实验表明，ReTAG在保持高响应质量的同时，显著降低了推理时间。代码已公开于https://github.com/bykimby/retag。
原文摘要 · Abstract (English)
Recent advances in question answering have led to substantial progress in tasks such as multi-hop reasoning. However, global sensemaking-answering questions by synthesizing information from an entire corpus remains a significant challenge. A prior graph-based approach to global sensemaking lacks retrieval mechanisms, topic specificity, and incurs high inference costs. To address these limitations, we propose ReTAG, a Retrieval-Enhanced, Topic-Augmented Graph framework that constructs topic-specific subgraphs and retrieves the relevant summaries for response generation. Experiments show that ReTAG improves response quality while significantly reducing inference time compared to the baseline. Our code is available at https://github.com/bykimby/retag.
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。