arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2404.07103cs.CLcs.IRcs.LG
收藏

图思维链:通过在图上的推理增强大语言模型

Graph Chain-of-Thought: Augmenting Large Language Models by Reasoning on Graphs

University of Illinois at Urbana-Champaign(伊利诺伊大学厄巴纳-香槟分校);Amazon(亚马逊公司);Pennsylvania State University(宾夕法尼亚州立大学);University of Virginia(弗吉尼亚大学)

Bowen Jin, Chulin Xie, Jiawei Zhang, Kashob Kumar Roy, Yu Zhang, Zheng Li, Ruirui Li, Xianfeng Tang, Suhang Wang, Yu Meng, Jiawei Han

AI总结:

本文提出图思维链(Graph-CoT)框架,通过让大语言模型在图数据上迭代推理来缓解幻觉问题,并构建GRBench基准验证其优于现有基线。

摘要:

大语言模型(LLMs)虽然表现出卓越的性能,但仍存在幻觉问题,尤其是在知识密集型任务上。现有工作提出利用从外部知识语料库中检索的单个文本单元来增强LLMs以缓解该问题。然而,在许多领域中,文本是相互关联的(例如,文献图谱中的学术论文通过引用和共同作者关系相互连接),从而形成(文本属性)图。此类图中的知识不仅编码在单个文本/节点中,还编码在它们之间的关联连接中。为促进利用图增强LLMs的研究,我们手动构建了一个名为GRBench的图推理基准数据集,包含1,740个问题,这些问题可借助来自10个领域图的知识来回答。随后,我们提出了一个简单有效的框架,称为图思维链(Graph-CoT),通过鼓励LLMs在图上迭代推理来利用图增强LLMs。每次Graph-CoT迭代包含三个子步骤:LLM推理、LLM-图交互和图执行。我们在GRBench上使用三种LLM主干进行了系统实验,Graph-CoT始终优于基线方法。代码可在https://github.com/PeterGriffinJin/Graph-CoT获取。

英文摘要:

Large language models (LLMs), while exhibiting exceptional performance, suffer from hallucinations, especially on knowledge-intensive tasks. Existing works propose to augment LLMs with individual text units retrieved from external knowledge corpora to alleviate the issue. However, in many domains, texts are interconnected (e.g., academic papers in a bibliographic graph are linked by citations and co-authorships) which form a (text-attributed) graph. The knowledge in such graphs is encoded not only in single texts/nodes but also in their associated connections. To facilitate the research of augmenting LLMs with graphs, we manually construct a Graph Reasoning Benchmark dataset called GRBench, containing 1,740 questions that can be answered with the knowledge from 10 domain graphs. Then, we propose a simple and effective framework called Graph Chain-of-thought (Graph-CoT) to augment LLMs with graphs by encouraging LLMs to reason on the graph iteratively. Each Graph-CoT iteration consists of three sub-steps: LLM reasoning, LLM-graph interaction, and graph execution. We conduct systematic experiments with three LLM backbones on GRBench, where Graph-CoT outperforms the baselines consistently. The code is available at https://github.com/PeterGriffinJin/Graph-CoT.

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。

↑