Reproduce “ComGPT: Detecting Local Community Structure with Large Language Models” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Li Ni, Haowen Shen, Lin Mu, Yiwen Zhang, Wenjian Luo
Researchers from Anhui University developed ComGPT, a GPT-guided algorithm for local community detection that leverages GPT-3.5-turbo's reasoning to overcome issues in traditional seed expansion methods. It employs a novel graph encoding method and specialized prompts, achieving superior performance on networks with distinct community structures such as Amazon and DBLP datasets.
Key insights
- Large Language Models can effectively reason about complex graph structures for local community detection when provided with specialized encoding and domain-aware prompts.
- Incorporating specific community-related knowledge directly into the graph encoding method (ComIncident) significantly improves the LLM's understanding and decision-making accuracy.
- Domain-specific prompt engineering (NSG and NSU) tailored to community detection principles outperforms generic advanced prompts, indicating the necessity of contextualized guidance for LLMs in specialized graph tasks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.