GraphInstruct

The GraphInstruct dataset is part of a benchmark proposed in the paper titled "GraphInstruct: Empowering Large Language Models with Graph Understanding and Reasoning Capability." This benchmark is designed to evaluate and enhance the graph understanding abilities of large language models (LLMs). It includes 21 classical graph reasoning tasks, providing diverse graph generation pipelines and detailed reasoning steps².

The dataset is used to construct models like GraphLM and GraphLM+, which are trained through efficient instruction-tuning and a step mask training strategy to show prominent graph understanding capability. These models have demonstrated superiority over other LLMs in understanding and reasoning with graph data².

(1) GraphInstruct: Empowering Large Language Models with Graph .... https://arxiv.org/abs/2403.04483. (2) CGCL-codes/GraphInstruct - GitHub. https://github.com/CGCL-codes/GraphInstruct. (3) GraphWiz/GraphInstruct · Datasets at Hugging Face. https://huggingface.co/datasets/GraphWiz/GraphInstruct/viewer. (4) GraphWiz: An Instruction-Following Language Model for Graph Problems. https://arxiv.org/abs/2402.16029. (5) undefined. https://doi.org/10.48550/arXiv.2403.04483.