README.md

metadata

license: apache-2.0
dataset_info:
  - config_name: human_object_interactions
    features:
      - name: video_id
        dtype: string
      - name: video_path
        dtype: string
      - name: question
        dtype: string
      - name: answer
        dtype: string
      - name: candidates
        dtype: string
      - name: source
        dtype: string
    splits:
      - name: mini
        num_bytes: 1749322
        num_examples: 5724
      - name: full
        num_bytes: 4074313
        num_examples: 14046
    download_size: 594056
    dataset_size: 5823635
  - config_name: intuitive_physics
    features:
      - name: video_id
        dtype: string
      - name: video_path
        dtype: string
      - name: question
        dtype: string
      - name: answer
        dtype: string
      - name: candidates
        dtype: string
      - name: source
        dtype: string
    splits:
      - name: mini
        num_bytes: 6999824
        num_examples: 8680
      - name: full
        num_bytes: 10570407
        num_examples: 12220
    download_size: 805372
    dataset_size: 17570231
  - config_name: robot_object_interactions
    features:
      - name: video_id
        dtype: string
      - name: video_path
        dtype: string
      - name: question
        dtype: string
      - name: answer
        dtype: string
      - name: candidates
        dtype: string
      - name: source
        dtype: string
    splits:
      - name: mini
        num_bytes: 771052
        num_examples: 2000
      - name: full
        num_bytes: 9954084
        num_examples: 25796
    download_size: 1547348
    dataset_size: 10725136
  - config_name: temporal_reasoning
    features:
      - name: video_id
        dtype: string
      - name: video_path
        dtype: string
      - name: question
        dtype: string
      - name: answer
        dtype: string
      - name: candidates
        dtype: string
      - name: source
        dtype: string
    splits:
      - name: mini
        num_bytes: 611241
        num_examples: 2012
      - name: full
        num_bytes: 783157
        num_examples: 2766
    download_size: 254438
    dataset_size: 1394398
configs:
  - config_name: human_object_interactions
    data_files:
      - split: mini
        path: human_object_interactions/mini-*
      - split: full
        path: human_object_interactions/full-*
  - config_name: intuitive_physics
    data_files:
      - split: mini
        path: intuitive_physics/mini-*
      - split: full
        path: intuitive_physics/full-*
  - config_name: robot_object_interactions
    data_files:
      - split: mini
        path: robot_object_interactions/mini-*
      - split: full
        path: robot_object_interactions/full-*
  - config_name: temporal_reasoning
    data_files:
      - split: mini
        path: temporal_reasoning/mini-*
      - split: full
        path: temporal_reasoning/full-*
task_categories:
  - question-answering
  - video-text-to-text
language:
  - en
size_categories:
  - 10K<n<100K

Minimal Video Pairs

A shortcut-aware benchmark for spatio-temporal and intuitive physics video understanding (VideoQA) using minimally different video pairs.

• Github

For legal reasons, we are unable to upload the videos directly to Huggingface. However, we provide scripts in this repository for downloading the videos in our github repository. Our benchmark is built on top of videos source from 9 domains:

Subset	Data sources
Human object interactions	PerceptionTest, SomethingSomethingV2
Robot object interactions	Language Table
Intuitive Physics and collisions	IntPhys, InfLevel, GRASP, CLEVRER
Temporal Reasoning	STAR, Vinoground

Run evaluation

To enable reproducible evaluation, we utilize the lmms-eval library. We have provided the task files you need to run mvp and mvp_mini. mvp_mini is essentially a smaller, balanced evaluation set with 9k examples for enabling faster evaluations.

Please follow the instructions on our Github repository for running reproducible evals.

Leaderboard

We also release our Physical Reasoning Leaderboard, where you can submit your model outputs for this dataset. Use the data_name as mvp and mvp_mini for submissions against full and mini split respectvely.

Citation and acknowledgements

We are grateful to the datasets listed above for utilizing their videos to create this benchmark. Please cite us if you use the benchmark or any other part of the paper:

@article{krojer2025shortcut,
  title={A Shortcut-aware Video-QA Benchmark for Physical Understanding via Minimal Video Pairs}
  author={Benno Krojer and Mojtaba Komeili and Candace Ross and Quentin Garrido and Koustuv Sinha and Nicolas Ballas and Mahmoud Assran},
  journal={arXiv},
  year={2025}
}