---
language:
- en
dataset_info:
  features:
  - name: id
    dtype: string
  - name: question
    dtype: string
  - name: options
    list: string
  - name: answer
    dtype: string
  - name: task_plan
    dtype: string
  - name: video
    dtype: binary
  splits:
  - name: video_3d_what_move
    num_bytes: 13156784
    num_examples: 408
  - name: video_3d_where_move
    num_bytes: 11911124
    num_examples: 402
  - name: video_3d_what_attribute_move
    num_bytes: 11401767
    num_examples: 390
  - name: video_3d_what_rotate
    num_bytes: 17223350
    num_examples: 411
  - name: video_3d_where_rotate
    num_bytes: 14860244
    num_examples: 378
  - name: video_3d_what_attribute_rotate
    num_bytes: 15306580
    num_examples: 405
  - name: video_sg_what_object
    num_bytes: 654650009
    num_examples: 387
  - name: video_sg_what_relation
    num_bytes: 707499456
    num_examples: 411
  - name: video_sg_what_action
    num_bytes: 636426745
    num_examples: 375
  download_size: 748458643
  dataset_size: 2082436059
configs:
- config_name: default
  data_files:
  - split: video_3d_what_move
    path: data/video_3d_what_move-*
  - split: video_3d_where_move
    path: data/video_3d_where_move-*
  - split: video_3d_what_attribute_move
    path: data/video_3d_what_attribute_move-*
  - split: video_3d_what_rotate
    path: data/video_3d_what_rotate-*
  - split: video_3d_where_rotate
    path: data/video_3d_where_rotate-*
  - split: video_3d_what_attribute_rotate
    path: data/video_3d_what_attribute_rotate-*
  - split: video_sg_what_object
    path: data/video_sg_what_object-*
  - split: video_sg_what_relation
    path: data/video_sg_what_relation-*
  - split: video_sg_what_action
    path: data/video_sg_what_action-*
---
# Dataset Card for TaskMeAnything-v1-videoqa-2024
<h2 align="center"> TaskMeAnything-v1-videoqa-2024 benchmark dataset</h2>

<h2 align="center"> <a href="https://www.task-me-anything.org/">🌐 Website</a> | <a href="https://arxiv.org/abs/2406.11775">📑 Paper</a> | <a href="https://huggingface.co/collections/jieyuz2/taskmeanything-664ebf028ab2524c0380526a">🤗 Huggingface</a> | <a href="https://huggingface.co/spaces/zixianma/TaskMeAnything-UI">💻 Interface</a></h2>
    
<h5 align="center"> If you like our project, please give us a star ⭐ on GitHub for latest update.  </h2>

## TaskMeAnything-v1-2024-Videoqa
[TaskMeAnything-v1-videoqa-2024](https://huggingface.co/datasets/weikaih/TaskMeAnything-v1-videoqa-2024) is a benchmark for reflecting the current progress of MLMs by `automatically` finding tasks that SOTA MLMs struggle with using the TaskMeAnything Top-K queries. 
This benchmark includes 2,394 3d video questions and 1,173 real video questions that the TaskMeAnything algorithm automatically approximated as challenging for over 12 popular MLMs.


The dataset contains 19 splits, while each splits contains 300+ questions from a specific task generator in TaskMeAnything-v1. For each row of dataset, it includes: video, question, options, answer and its corresponding task plan.

## Load TaskMeAnything-v1-2024 VideoQA Dataset
```
import datasets

dataset_name = 'weikaih/TaskMeAnything-v1-videoqa-2024'
dataset = datasets.load_dataset(dataset_name, split = TASK_GENERATOR_SPLIT)
```
where `TASK_GENERATOR_SPLIT` is one of the task generators, eg, `2024_2d_how_many`.

## Evaluation Results

### Overall


![image/png](https://cdn-uploads.huggingface.co/production/uploads/65cb0dcc4913057ac82a7a31/7QII8uTpDc5YwnRmKVfZH.png)

### Breakdown performance on each task types

![image/png](https://cdn-uploads.huggingface.co/production/uploads/65cb0dcc4913057ac82a7a31/WKMcravDMljrpqvtlVGzq.png)

![image/png](https://cdn-uploads.huggingface.co/production/uploads/65cb0dcc4913057ac82a7a31/mFV2GcyxsVEFhEM9rBwWb.png)

## Out-of-Scope Use
This dataset should not be used for training models.


## Disclaimers
**TaskMeAnything** and its associated resources are provided for research and educational purposes only. 
The authors and contributors make no warranties regarding the accuracy or reliability of the data and software. 
Users are responsible for ensuring their use complies with applicable laws and regulations. 
The project is not liable for any damages or losses resulting from the use of these resources.

## Contact

- Jieyu Zhang: jieyuz2@cs.washington.edu

## Citation
**BibTeX:**
```bibtex
@article{zhang2024task,
  title={Task Me Anything},
  author={Zhang, Jieyu and Huang, Weikai and Ma, Zixian and Michel, Oscar and He, Dong and Gupta, Tanmay and Ma, Wei-Chiu and Farhadi, Ali and Kembhavi, Aniruddha and Krishna, Ranjay},
  journal={arXiv preprint arXiv:2406.11775},
  year={2024}
}
```