Datasets and Experimental Outputs for Integrating Large Language Models and Graph Convolutional Networks for Semi-Supervised Image Classification

This repository contains the datasets and intermediate files required to reproduce the experiments presented in the accompanying paper. The source code is available on GitHub, while large datasets and generated files are hosted on Zenodo due to GitHub storage limitations. The archive includes: Corel5K image dataset visual feature representations image ranking files automatically generated captions (BLIP) caption-ranking files LLM similarity scores The provided files allow reproducing the experiments without rerunning computationally expensive preprocessing and LLM inference stages.