DualView: Adaptive Local-Global Fusion for Multi-Hop Document Reranking
The paper presents a new framework called DualView for multi-hop document reranking, which is essential for effective question answering. It combines local and global scoring mechanisms to enhance the identification of relevant documents while maintaining high recall. The model demonstrates superior performance compared to existing methods, achieving high accuracy and low latency.
- ▪DualView employs a Local Scorer for fine-grained query-document relevance using stacked cross-attention.
- ▪A Global Scorer models inter-document dependencies through Transformer-based context aggregation.
- ▪The model achieves 99.4% Top-4 Recall and 97.8% Full Hit accuracy at a latency of 4.0 ms, outperforming larger cross-encoders.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Information Retrieval arXiv:2605.18767 (cs) [Submitted on 13 Apr 2026] Title:DualView: Adaptive Local-Global Fusion for Multi-Hop Document Reranking Authors:Litong Zhang, Jiaxin Li, Kuo Zhao View a PDF of the paper titled DualView: Adaptive Local-Global Fusion for Multi-Hop Document Reranking, by Litong Zhang and 2 other authors View PDF HTML (experimental) Abstract: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.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
