BAAI/JudgeLM-data-collection-v1.0
收藏资源简介:
该数据集名为JudgeLM-data-collection-v1.0,旨在为评估JudgeLM模型提供便利。它包含了由大型语言模型生成的答案以及一个名为MM-Vet的多模态基准测试。数据集的文件结构包括多个JSONL文件,这些文件可能包含模型的输出、预处理脚本和参考数据。数据集的语言为英语,遵循CC BY-NC 4.0许可。
This dataset, titled JudgeLM-data-collection-v1.0, is developed to facilitate the evaluation of the JudgeLM model. It encompasses answers generated by large language models and a multimodal benchmark named MM-Vet. The file structure of the dataset comprises multiple JSONL files, which may contain model outputs, preprocessing scripts, and reference data. The dataset is in English and is released under the CC BY-NC 4.0 license.
数据集卡片 for JudgeLM-data-collection
数据集描述
数据集概述
该数据集旨在方便使用和评估 JudgeLM。我们包含了 LLMs 生成的答案和一个多模态基准 MM-Vet。文件夹结构如下:
文件夹结构
data ├── JudgeLM/ │ ├── answers/ │ │ ├── alpaca_judgelm_val.jsonl | | ├── ... │ ├── judgelm_preprocess.py │ ├── judgelm_val_5k.jsonl │ ├── judgelm_val_5k_gpt4.jsonl │ ├── judgelm_val_5k_gpt4_with_reference.jsonl │ ├── judgelm_val_5k_references.jsonl ├── MM-Vet/ │ ├── mm-vet-emu-prediction.json │ ├── mm-vet-gt.json │ ├── mm-vet-judge-samples.jsonl │ ├── mmvet_preprocess.py
您可以直接将此数据集集合放入 /JudgeLM/judgelm 中以更好地使用。
语言
Alpaca 中的数据为英语。
附加信息
许可信息
该数据集在 Creative Commons NonCommercial (CC BY-NC 4.0) 许可下可用。
引用信息
@article{zhu2023judgelm,
title={JudgeLM: Fine-tuned Large Language Models are Scalable Judges},
author={Lianghui Zhu and Xinggang Wang and Xinlong Wang},
year={2023},
eprint={2310.17631},
archivePrefix={arXiv},
primaryClass={cs.CL}
}




