Conceptual
Browse Instruct Curate

Graph2text versus Graph2token Taxonomy for LLM-Based Graph Learning

A taxonomy (LLM4graph) that organizes methods for applying large language models to graphs by how the graph is turned into model input. Graph2text serializes the graph into natural-language text the LLM reads; Graph2token encodes the graph into token embeddings fed directly to the model. The framework names four transformation challenges, uses them to sort existing methods problem-first, and provides guidance for choosing a method under given graph and hardware constraints.