通过动态融合局部与全局信息，高效提升多跳文档重排序效果。
DualView: Adaptive Local-Global Fusion for Multi-Hop Document Reranking
- 双视角架构：局部细粒度匹配+全局文档依赖建模
- 在MuSiQue数据集上达99.4%顶4召回率，延迟仅4毫秒
- 适合追求高精度与低延迟的问答系统研发人员
多跳问答需整合多篇文档信息，关键挑战在于从候选文档中高效识别最小相关集合的同时保持高召回率。本文提出一种轻量级双视角级联重排序框架，作用于E5-base-v2的检索结果。模型包含：(1) 基于堆叠交叉注意力的局部评分器，实现细粒度查询-文档相关性判断；(2) 基于Transformer的上下文聚合全局评分器，建模文档间依赖关系。二者通过依据查询语义动态调整的自适应门控融合。在固定候选集重排序设定下，使用离线缓存嵌入，模型在MuSiQue数据集上达到99.4%的Top-4 Recall和97.8%的Full Hit准确率，延迟仅4.0毫秒（249 QPS），显著优于600M参数的交叉编码器（BGE-Large: 92.0% Recall, Jina-v3: 90.1% Recall），且延迟降低5至6倍。消融实验验证了局部与全局视角对多跳性能均有显著贡献。
原文摘要 · Abstract (English)
Multi-hop question answering requires aggregating information from multiple documents, a critical capability for knowledge-intensive applications. A fundamental challenge lies in efficiently identifying the minimal relevant document set from retrieved candidates while maintaining high recall. We present an efficient dual-view cascaded reranking framework for multi-hop document reranking. Operating as a lightweight post-retrieval stage over E5-base-v2 candidates, our architecture comprises: (1) a Local Scorer employing stacked cross-attention for fine-grained query-document relevance; and (2) a Global Scorer modeling inter-document dependencies via Transformer-based context aggregation. These views are dynamically fused through an Adaptive Gate conditioned on query semantics. Under the fixed candidate set reranking setting with offline cached embeddings, our model achieves competitive results, particularly outstanding on MuSiQue with 99.4% Top-4 Recall and 97.8% Full Hit accuracy at 4.0 ms latency (249 QPS). It substantially outperforms 600M-parameter cross-encoders (BGE-Large: 92.0% Recall, Jina-v3: 90.1% Recall) while maintaining 5 to 6 times lower latency. Ablation studies validate that both Local and Global views contribute substantially to multi-hop performance.
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。