Datasets:
Tasks:
Image-Text-to-Text
Formats:
parquet
Languages:
English
Size:
< 1K
ArXiv:
Tags:
image-text-to-text
License:
metadata
language:
- en
license: apache-2.0
task_categories:
- image-text-to-text
tags:
- image-text-to-text
configs:
- config_name: default
data_files:
- split: test
path: vstar.parquet
The evaluation data for the V* benchmark.
The data structure follows the evaluation code of PixelReasoner