GraphGPT and HiGPT
GraphGPT focuses on instruction-tuning LLMs for general graph understanding tasks, while HiGPT specializes in heterogeneous graphs with multiple node and edge types, making them complementary approaches that address different graph structures rather than competing solutions.
About GraphGPT
HKUDS/GraphGPT
[SIGIR'2024] "GraphGPT: Graph Instruction Tuning for Large Language Models"
This project helps researchers and data scientists working with complex graph data to integrate that data more effectively with large language models (LLMs). It takes graph-structured data (like networks of academic papers or biological interactions) and processes it so LLMs can 'understand' and reason about its connections. The output is a fine-tuned language model capable of performing tasks like node classification or link prediction within the graph context, making it useful for researchers analyzing networked information.
About HiGPT
HKUDS/HiGPT
[KDD'2024] "HiGPT: Heterogenous Graph Language Models"
This project helps people understand complex relationships within diverse, interconnected data, often found in fields like academic networks or movie databases. By inputting structured data that describes entities and their various connections (e.g., authors, papers, venues), it generates insights or answers questions based on these relationships, going beyond simple keyword matching. Researchers, data scientists, or analysts working with intricate network data would find this useful.
Scores updated daily from GitHub, PyPI, and npm data. How scores work