---
language:
- en
license: mit
size_categories:
- n<1K
tags:
- vision-language
- reward-modeling
- benchmark
task_categories:
- image-text-to-text
---

Benchmark Data for ViLBench: A Suite for Vision-Language Process Reward Modeling

[arXiv](https://arxiv.org/abs/2503.20271) | [Project Page](https://ucsc-vlaa.github.io/ViLBench/)

There are 600 data collected from 5 existing vision-language tasks