search

HKUDS / GraphGPT

[SIGIR'2024] "GraphGPT: Graph Instruction Tuning for Large Language Models"
https://arxiv.org/abs/2310.13023
Apache License 2.0
576 stars 53 forks source link

Can the GraphGPT be used for LLM inference on other domain graph without retrainng? #25

Closed lushunn closed 10 months ago

lushunn commented 10 months ago

It's a wonderful research, but I have a question: all the training data seems to be collected from the research paper domain. Therefore, if it lacks the capability to zero-shot infer graphs from other domains, such as medicine or communication ?

tjb-tech commented 10 months ago

It's a wonderful research, but I have a question: all the training data seems to be collected from the research paper domain. Therefore, if it lacks the capability to zero-shot infer graphs from other domains, such as medicine or communication ?

Thank you for your interest in our GraphGPT. I apologize for the delayed response due to the academic workload at the end of the semester. Your question is great! I think there are two ways to find out the answer:

  1. You can directly run GraphGPT using your domain-specific data and evalute it.
  2. If the results in your domain-specific data are not that good. I recommend you to tune GraphGPT in your data.

And we believe that the zero-shot ability of GraphGPT in other domain graphs is still good. Hope my answer is help for you. Wishing you an early Merry Christmas!