通过合成的分步监督增强代理文本图检索
Enhancing Agentic Textual Graph Retrieval with Synthetic Stepwise Supervision
AI总结：
本文提出一种基于LLM的文本图检索框架，通过合成分步监督优化检索过程，提升复杂图基问答任务的准确性和F1分数。
摘要：
将文本图整合进大型语言模型（LLMs）对于复杂图基问答任务具有潜力。然而，检索具有信息性且紧凑的子图以适应LLM上下文是关键瓶颈。现有检索器常 struggles，依赖于浅层嵌入相似性或成本高昂的交互策略，需要过多监督。为解决这些问题，我们引入了一个代理文本图推理框架，其特征是通过合成分步监督训练的LLM基于检索器。而不是依赖最终答案奖励，通常产生稀疏且不稳定的信号，我们通过评估每一步与离线提取的黄金子图进行优化。我们的方法通过专门的数据合成管道提炼黄金子图，形成密集奖励，从而实现一个两阶段训练方案，有效学习交互图探索策略。基于在三个常见数据集上的广泛实验，与七个强大的基线相比，我们的方法在准确率上平均提高了15.6%，F1分数提高了17.2%。在更复杂的多跳推理任务中，优势更高。
英文摘要：
Integrating textual graphs into Large Language Models (LLMs) is promising for complex graph-based QA. However, a key bottleneck is retrieving informative yet compact subgraphs that fit the LLM context. Existing retrievers often struggle, relying either on shallow embedding similarity or costly interactive policies that require excessive supervision. To address these challenges, we introduce an agentic textual graph reasoning framework featuring an LLM-based retriever trained with synthetic stepwise supervision. Rather than relying on final answer rewards which often yield sparse and unstable signals, we optimize the retriever by evaluating each step against offline-extracted golden subgraphs. Our approach distills golden subgraphs via a specialized data synthesis pipeline to formulate dense rewards, facilitating a two-stage training scheme that effectively learns the interactive graph exploration policy. Based on extensive experiments on three common datasets in comparison with seven strong baselines, our approach achieves an average improvement of 15.6% in accuracy and 17.2% in F1 score. The advantage is even higher in more complicated multi-hop reasoning tasks.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
