Datasets:
| 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 |