This project, wilI intend to propose graph problem-solving tasks for different models, taking the BIG-benchmark format as a starting point and using models such as BLOOM, GPT-3, GPT-4, T5, FLAN-T5, and BERT.