Molmo2-ER-SAT / README.md
Duanj1's picture
Initial release: array/SAT subset for Molmo2-ER training
421ec9e verified
metadata
license: mit
language:
  - en
pretty_name: Molmo2-ER SAT
tags:
  - embodied-reasoning
  - molmo2
  - molmo2-er
  - vlm-training-data

Molmo2-ER · array/SAT

Spatial Aptitude Training: 175K binary-MCQ VQA pairs over ProcTHOR indoor scenes.

This is a re-hosted, loader-ready subset of the upstream dataset, used to train allenai/Molmo2-ER-4B. Files mirror the upstream layout; nothing in the data has been modified.

Upstream source

  • Original dataset: array/SAT
  • Paper: SAT: Dynamic Spatial Aptitude Training for Multimodal Language Models (arXiv:2412.07755)
  • License: mit (inherits from upstream)

If you use this data, please cite the original authors:

@misc{ray2025satdynamicspatialaptitude,
  title={SAT: Dynamic Spatial Aptitude Training for Multimodal Language Models},
  author={Arijit Ray and Jiafei Duan and Ellis Brown and others},
  year={2025},
  eprint={2412.07755},
  archivePrefix={arXiv}
}

Usage in Molmo2-ER

See the allenai/molmo2 repository for the data loader and training recipe. The relevant loader class for this dataset lives in olmo/data/spatial_datasets.py.