Avaliação - MME-Perception
Estrutura do Diretório
main
├── MME_Benchmark
│ ├── artwork
│ │ ├── images
│ │ │ ├── 1.jpg
│ │ │ ├── 2.jpg
│ │ │ ├── ...
│ │ ├── question_answers_YN
│ │ │ ├── 1.txt
│ │ │ ├── 2.txt
│ │ │ ├── ...
│ ├── celebrity
│ ├── code_reasoning
│ ├── ...
├── calculation.py
├── translate_MME
Cada arquivo num.txt contém as perguntas correspondentes à imagem num.jpg.… See the full description on the dataset page:
https://huggingface.co/datasets/LucasLima/MME-Benchmark-pt.