ParallaxRAG：基于多视图知识图谱 RAG 解决多跳推理问题
Think Parallax: Solving Multi-Hop Problems via Multi-View Knowledge-Graph-Based Retrieval-Augmented Generation
AI 导读
论文提出 ParallaxRAG，一种对称多视图框架，将查询与知识图谱解耦到与注意力头对齐的语义空间，构建更干净的子图，引导 LLM 进行逐跳推理。该框架在 WebQSP 和 CWQ 上取得 SOTA 的检索与问答表现，显著减少模型幻觉，并在生物医学基准 BioASQ 上展现出良好泛化能力。
正文 · 原文
Abstract:Large language models (LLMs) still struggle with multi-hop reasoning over knowledge-graphs (KGs), and we identify a previously overlooked structural reason for this difficulty: Transformer attention heads naturally specialize in distinct semantic relations across reasoning stages, forming a hop-aligned relay pattern. This key finding suggests that multi-hop reasoning is inherently multi-view, yet existing KG-based retrieval-augmented generation (KG-RAG) systems collapse all reasoning hops into a single representation, flat embedding space, suppressing this implicit structure and causing noisy or drifted path exploration. We introduce ParallaxRAG, a symmetric multi-view framework that decouples queries and KGs into aligned, head-specific semantic spaces. By enforcing relational diversity across multiple heads while constraining weakly related paths, ParallaxRAG constructs more accurate, cleaner subgraphs and guides LLMs through grounded, hop-wise reasoning. On WebQSP and CWQ, it achieves state-of-the-art retrieval and QA performance, substantially reduces hallucination, and generalizes strongly to the biomedical BioASQ benchmark.
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2510.15552 [cs.CL] |
| (or arXiv:2510.15552v5 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2510.15552 arXiv-issued DOI via DataCite |
Submission history
From: Jinliang Liu [view email]
[v1]
Fri, 17 Oct 2025 11:34:27 UTC (3,364 KB)
[v2]
Mon, 29 Dec 2025 07:32:34 UTC (3,367 KB)
[v3]
Mon, 13 Apr 2026 09:34:10 UTC (3,132 KB)
[v4]
Tue, 14 Apr 2026 01:45:11 UTC (3,132 KB)
[v5]
Mon, 26 Oct 2026 03:16:11 UTC (3,488 KB)
来源：arXiv cs.AI · arxiv.org