search

wjn1996 / InstructGraph

A framework to empover LLMs on graph reasoning and generation. Refer to our paper: https://arxiv.org/pdf/2402.08785.pdf
MIT License
72 stars 4 forks source link

May add another related work as reference #10

Closed WEIYanbin1999 closed 8 months ago

WEIYanbin1999 commented 8 months ago

Dear Author,

During reading your paper, I noticed that our research work, "KICGPT: Large Language Model with Knowledge in Context for Knowledge Graph Completion," which was accepted to EMNLP Findings in 2023, is highly relevant to your field of study. KICGPT focuses on enhancing the completeness of knowledge graphs by utilizing context knowledge, and can be categorized into works that utilize prompt engineering or RAG.

I believe incorporating our work into your list of references will add depth to your study while also offering a more comprehensive knowledge base for scholars in the related fields. If you find it appropriate, I would greatly appreciate it if you could cite our paper.

Wishing you smooth progress in your research!

wjn1996 commented 8 months ago

Dear researcher,

I appreciate your reminder. I have read your paper and I think the proposed KICGPT is similar to prompt engineering (e.g. ICL), which designs demonstrations and templates to prompt the LLM to generate missing components in the triple and then uses self-alignment to perform post-processing. Therefore, your suggestion can be accepted and your paper can be included in the reference list for the next version.

However, as your work only focuses on knowledge graph completion rather than generating the entire knowledge graph or reasoning on the graph, I think the KICGPT may not be appropriate as our baseline."

Please let me know if you have any questions.

Best, Jianing.

WEIYanbin1999 commented 8 months ago

Thank you for your positive response. Best.