LADDER: Graph-Guided Diffusion Language Models for Efficient Multi-Hop Reasoning

📄 arXiv: 2609.24346v1 📥 PDF

作者: Senlei Zhang, Linhao Luo, Qian-Wen Zhang, Siyu An, Junnan Dong, Shuhao Zhang, Xing Sun

分类: cs.AI

发布日期: 2026-09-21

💡 一句话要点

提出LADDER以解决多跳推理效率低下问题

🎯 匹配领域: 支柱九：具身大模型 (Embodied Foundation Models)

关键词: 图检索增强生成 多跳推理 扩散语言模型 并行解码 知识图谱

📋 核心要点

  1. 现有的GraphRAG框架依赖自回归语言模型，导致推理效率低下，特别是在多跳推理任务中。
  2. LADDER通过图引导的并行解码，采用事件驱动的自时钟检索和不完整查询图传播模块，解决了动态图基础上的推理问题。
  3. 在三个多跳问答基准上，LADDER将准确率从39.6%提升至45.2%，并实现了4.1倍的延迟减少，显示出显著的性能提升。

📝 摘要（中文）

图检索增强生成（GraphRAG）通过利用结构化实体拓扑显著提升了大型语言模型在复杂推理任务中的表现。然而，现有框架依赖于标准的自回归语言模型，其固有的序列生成特性严重制约了推理效率。本文提出LADDER，一个新颖的框架，通过图引导的并行解码将扩散语言建模与GraphRAG结合，旨在加速离散空间中的推理过程。我们提出了事件驱动的自时钟检索机制和不完整查询图传播模块，显著提高了推理的准确性和效率。实验结果表明，LADDER在多跳问答基准上将平均准确率从39.6%提升至45.2%，同时实现了4.1倍的延迟减少。

🔬 方法详解

问题定义：本文旨在解决现有GraphRAG框架在多跳推理中效率低下的问题。现有方法依赖于自回归生成，导致推理过程缓慢且不稳定。

核心思路：LADDER通过结合扩散语言模型的并行解码能力与图检索增强生成，提出了一种新的推理框架，旨在提高推理效率和准确性。

技术框架：LADDER的整体架构包括事件驱动的自时钟检索模块和不完整查询图传播模块。自时钟检索模块在部分去噪状态下动态触发图检索，而图传播模块则聚合多跳证据以增强并行预测。

关键创新：LADDER的主要创新在于引入了事件驱动的自时钟检索机制，能够在目标实体早期出现时快速响应，从而显著减少推理步骤。此外，不完整查询图传播模块有效处理新出现的实体查询，提升了推理的准确性。

关键设计：在设计中，LADDER采用了动态触发的图检索策略，避免了传统方法中的学习门控或启发式阈值，同时使用了专门的图基础模型来处理新实体查询，确保了多跳证据的有效聚合。

🖼️ 关键图片

fig_0
img_1
img_2

📊 实验亮点

LADDER在三个多跳问答基准上实现了显著的性能提升，平均准确率从39.6%提高至45.2%，同时延迟减少了4.1倍，展示了其在复杂推理任务中的有效性和高效性。

🎯 应用场景

LADDER框架在多跳推理任务中展现出优越的性能，具有广泛的应用潜力。其可用于复杂的问答系统、知识图谱推理以及其他需要高效推理的自然语言处理任务，未来可能推动智能助手和自动化决策系统的发展。

📄 摘要（原文）

Graph Retrieval-Augmented Generation (GraphRAG) has remarkably enhanced large language models on complex reasoning by leveraging structured entity topologies. However, existing frameworks heavily rely on standard autoregressive language models where the nature of inherent sequential generation severely hinders overall inference efficiency. Inspired by Diffusion Language Models (DLMs) that offer massive parallelism via continuous refine-in-parallel decoding, we aim to accelerate GraphRAG in the discrete space. However, it remains non-trivial for two challenges. First, partially denoised drafts are highly dynamic and uncertain, making dynamic graph grounding non-trivial. Second, raw denoising states are inherently noisy and unstable, making synchronous graph retrieval and multi-hop aggregation computationally prohibitive. To this end, we present LADDER, a novel framework that bridges diffusion language modeling with GraphRAG through graph-guided parallel decoding. Specifically, (i) we propose an event-driven self-clocking retrieval, inspired by our key insight that 88% of target entities emerge early in the partially denoised state, leading final commitment by an average of 5.7-9.6 steps. This mechanism dynamically triggers graph retrieval only when the set of graph-linkable entities expands, yielding an asynchronous self-clocking policy that bypasses learned gates or heuristic thresholds. (ii) An incomplete-query graph propagation module is designed to process the newly emerging entity queries using a specialized graph foundation model, continuously aggregating multi-hop evidence to sharpen parallel predictions and accelerate overall decoding convergence. Extensive experiments on three challenging multi-hop QA benchmarks show that LADDER raises average exact match from 39.6% to 45.2% while achieving a 4.1x latency reduction.