zjunlp/AutoKG
Framework for evaluating LLMs on knowledge graph construction and reasoning tasks using multi-agent systems.
Not currently ranked — collecting fresh signals.
star history
This repository accompanies a WWW 2024 paper assessing large language models’ capabilities for knowledge graph construction and reasoning. It evaluates models including ChatGPT, GPT-4, and text-davinci-003 on benchmarks like RE-TACRED, DuIE2.0, and MAVEN for relation extraction and event extraction. The work also proposes using multiple agents to autonomously construct and reason over knowledge graphs, along with analyzing virtual knowledge extraction capabilities via the VINE dataset.
Frequently asked
- What is zjunlp/AutoKG?
- Framework for evaluating LLMs on knowledge graph construction and reasoning tasks using multi-agent systems.
- Is AutoKG open source?
- Yes — zjunlp/AutoKG is open source, released under the MIT license.
- What language is AutoKG written in?
- zjunlp/AutoKG is primarily written in Python.
- How popular is AutoKG?
- zjunlp/AutoKG has 469 stars on GitHub.
- Where can I find AutoKG?
- zjunlp/AutoKG is on GitHub at https://github.com/zjunlp/AutoKG.