ExpVid

# ExpVid: A Benchmark for Experiment Video Understanding & Reasoning
We present **ExpVid**, a benchmark to evaluate MLLMs on scientific experiment videos. ExpVid comprises 10 tasks across 3 levels, curated from a collection of 390 lab experiment videos spanning 13 disciplines.


# How to Use
```python
from datasets import load_dataset
dataset = load_dataset("OpenGVLab/ExpVid")
```

All task annotation `.jsonl` files are stored under `annotations/level_*`.
- Each annotation includes the field:
- `video_path`: the relative path pointing to the video corresponding to that QA.
- Other common fields: `question`, `options`, `answer`, `asr_caption`, etc.

# Citation
If you find this work useful for your research, please consider citing ExpVid. Your acknowledgement would greatly help us in continuing to contribute resources to the research community.
[https://arxiv.org/abs/2510.11606](https://arxiv.org/abs/2510.11606)
```bibtex
@article{xu2025expvid,
title={ExpVid: A Benchmark for Experiment Video Understanding \& Reasoning},
author={Xu, Yicheng and Wu, Yue and Yu, Jiashuo and Yan, Ziang and Jiang, Tianxiang and He, Yinan and Zhao, Qingsong and Chen, Kai and Qiao, Yu and Wang, Limin and Okumura, Manabu and Wang, Yi},
journal={arXiv preprint arXiv:2510.11606},
year={2025}
}

# ExpVid：面向实验视频理解与推理的基准数据集
本次工作提出**ExpVid**，一款用于评估多模态大语言模型（Multimodal Large Language Model，MLLM）科学实验视频理解与推理能力的基准数据集。ExpVid由覆盖13个学科的390个实验室实验视频构建而来，包含3个难度层级下的10项任务。


# 使用方法
python
from datasets import load_dataset
dataset = load_dataset("OpenGVLab/ExpVid")

所有任务标注的`.jsonl`文件均存储于`annotations/level_*`路径下。
- 每条标注均包含以下字段：
- `video_path`：指向当前问答对对应视频的相对路径。
其余通用字段包括：`question`（问题）、`options`（选项）、`answer`（答案）、`asr_caption`（语音转文字字幕）等。

# 引用声明
若本数据集对您的研究有所助益，请引用ExpVid。您的认可将极大助力我们持续为学术社区贡献相关资源。
[https://arxiv.org/abs/2510.11606](https://arxiv.org/abs/2510.11606)
bibtex
@article{xu2025expvid,
title={ExpVid: A Benchmark for Experiment Video Understanding & Reasoning},
author={Xu, Yicheng and Wu, Yue and Yu, Jiashuo and Yan, Ziang and Jiang, Tianxiang and He, Yinan and Zhao, Qingsong and Chen, Kai and Qiao, Yu and Wang, Limin and Okumura, Manabu and Wang, Yi},
journal={arXiv preprint arXiv:2510.11606},
year={2025}
}