VisualPRM400K

# VisualPRM400K

[\[📂 GitHub\]](https://github.com/OpenGVLab/InternVL)
[\[📜 Paper\]](https://arxiv.org/abs/2503.10291)
[\[🆕 Blog\]](https://internvl.github.io/blog/2025-03-13-VisualPRM/)
[\[🤗 model\]](https://huggingface.co/OpenGVLab/VisualPRM-8B)
[\[🤗 dataset\]](https://huggingface.co/datasets/OpenGVLab/VisualPRM400K)
[\[🤗 benchmark\]](https://huggingface.co/datasets/OpenGVLab/VisualProcessBench)

***`2025/04/11:` We release a new version of VisualPRM400K (i.e., [VisualPRM400K-v1.1](https://huggingface.co/datasets/OpenGVLab/VisualPRM400K-v1.1)), which includes additional data sources to enhance the data diversity.***

VisualPRM400K is a dataset comprising approximately 400K multimodal process supervision data. We generate the data using an automatic data pipeline. The key idea is to estimate the expected accuracy \\(mc_i\\) of the given step \\(s_{\leq i}\\) based on Monte Carlo sampling and consider the step correct if \\(mc_i>0\\). Please see our [paper](https://arxiv.org/abs/2503.10291) or [blog](https://internvl.github.io/blog/2025-03-13-VisualPRM/) for more details.

NOTE: This dataset is formulated as multi-turn conversation and the expected accuracy \\(mc_i\\) has been converted into correctness token \\(c_i \in \{+,-\}\\). If you want to use the annotations for expected accuracy, please refer to [this version](https://huggingface.co/datasets/OpenGVLab/VisualPRM400K-v1.1-Raw).

## Data Examples


## License

This project is released under the MIT License. This project uses the pre-trained internlm2_5-7b-chat as a component, which is licensed under the Apache License 2.0.

## Citation

If you find this project useful in your research, please consider citing:

```BibTeX
@article{wang2025visualprm,
title={VisualPRM: An Effective Process Reward Model for Multimodal Reasoning},
author={Wang, Weiyun and Gao, Zhangwei and Chen, Lianjie and Chen, Zhe and Zhu, Jinguo and Zhao, Xiangyu and Liu, Yangzhou and Cao, Yue and Ye, Shenglong and Zhu, Xizhou and others},
journal={arXiv preprint arXiv:2503.10291},
year={2025}
}
```

# VisualPRM400K
[📂 GitHub 仓库](https://github.com/OpenGVLab/InternVL)
[📜 学术论文](https://arxiv.org/abs/2503.10291)
[🆕 官方博客](https://internvl.github.io/blog/2025-03-13-VisualPRM/)
[🤗 模型权重](https://huggingface.co/OpenGVLab/VisualPRM-8B)
[🤗 数据集资源](https://huggingface.co/datasets/OpenGVLab/VisualPRM400K)
[🤗 基准测试集](https://huggingface.co/datasets/OpenGVLab/VisualProcessBench)

***2025/04/11：*** 我们发布了 VisualPRM400K 的更新版本（即 [VisualPRM400K-v1.1](https://huggingface.co/datasets/OpenGVLab/VisualPRM400K-v1.1)），该版本新增多类数据源以提升数据多样性。

VisualPRM400K 是包含约40万条多模态过程监督数据的数据集，我们通过自动化数据流水线生成该数据集。其核心思路是基于蒙特卡洛（Monte Carlo）采样，估算给定步骤 \(s_{\leq i}\) 的预期准确率 \(mc_i\)，当 \(mc_i>0\) 时即认为该步骤正确。如需了解更多细节，请参阅我们的[学术论文](https://arxiv.org/abs/2503.10291)或[官方博客](https://internvl.github.io/blog/2025-03-13-VisualPRM/)。

**注意：** 本数据集采用多轮对话格式组织，且预期准确率 \(mc_i\) 已被转换为正确性标记 \(c_i \in \{+, -\}\)。若需使用预期准确率的原始标注，请参阅[该版本数据集](https://huggingface.co/datasets/OpenGVLab/VisualPRM400K-v1.1-Raw)。

## 数据示例


## 许可证

本项目采用 MIT 许可证开源。本项目使用预训练模型 internlm2_5-7b-chat 作为组件，该模型遵循 Apache 许可证2.0开源。

## 引用格式

若本项目对您的研究有所帮助，请考虑引用如下文献：

BibTeX
@article{wang2025visualprm,
title={VisualPRM: An Effective Process Reward Model for Multimodal Reasoning},
author={Wang, Weiyun and Gao, Zhangwei and Chen, Lianjie and Chen, Zhe and Zhu, Jinguo and Zhao, Xiangyu and Liu, Yangzhou and Cao, Yue and Ye, Shenglong and Zhu, Xizhou and others},
journal={arXiv preprint arXiv:2503.10291},
year={2025}
}