metadata
language:
- en
tags:
- sparql
- rdf
- knowledge-graph
- large-language-models
- llm
- llm-benchmarking
pretty_name: LLM-KG-Bench Leaderboard
configs:
- config_name: combined_scores_short
data_files: board_combined_scores_short.csv
default: true
- config_name: combined_scores
data_files: board_combined_scores.csv
- config_name: main_scores
data_files: board_main_scores.csv
- config_name: all_scores
data_files: board_all_stats.csv
Leaderboard for RDF Knowledge Graph(KG) related capabilities of Large Language Models(LLMs) as generated with the LLM-KG-Bench framework.
Results for more than 20 RDF related tasks are collected for more than 40 LLMs. The leaderboard contains summarized results:
- board_combined_scores_short.csv: the most concise summary, listing for each LLM combined scores in the RDF (R) and SPARQL (S) handling categories, estimating read(R) and write(W) capabilities for syntax(syn) and semantic(sem). If a dialogue was tested, the result for the first(1) and best(max) answer is given. board_combined_scores.csv is optimized for automated parsing.
- board_main_scores.csv contains the summarized stats for the most important scores
- board_all_stats.csv contains the summarized stats for all scores of all tasks
The full leaderboard with more explanation can be found at https://aksw.github.io/LLM-KG-Bench-Leaderboard/