VLA-OS: Training Data for Vision-Language-Action Model Planning Representations

A training dataset used in the paper 'VLA-OS: Structuring and Dissecting Planning Representations and Paradigms in Vision-Language-Action Models'. The dataset was created by Linslab and the associated paper was published on arXiv in June 2025. It is hosted on Hugging Face and was last updated on 2025-06-24.

Use Cases

• Training vision-language-action models based on the planning paradigms described in the paper.
• Benchmarking model performance on structured planning tasks based on the dataset's intended use.
• Analyzing the effectiveness of different planning representations in multimodal AI systems based on the research focus.

Strengths

• Dataset is directly associated with a peer-reviewed research paper (arXiv:2506.17561).
• Provides a direct link to the source code and model for reproducibility.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and dataset size are unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Linslab, associated with the National University of Singapore (NUS) Lins Lab.

Collection Method

Created as the training dataset for the VLA-OS research project.

Freshness

Last updated 2025-06-24 04:00:27.