VisionRewardDB Image Regression: Paired Images with Metadata

Visionrewarddb Image Regression is a dataset containing pairs of images for regression tasks, created by zai-org. The dataset structure includes a metadata file (regression.jsonl) linking image pairs with source tags and a separate compressed archive for images from the ImageRewardDB source. It was last updated on Hugging Face on July 2, -2025.

Use Cases

• Training regression models to predict scores based on image pairs mentioned in the description
• Benchmarking image quality or aesthetic assessment algorithms using paired image data
• Developing or evaluating vision-language reward models based on the image pair structure

Strengths

• Dataset structure is clearly documented with a dedicated metadata file (regression.jsonl)
• Image data from the ImageRewardDB source is provided in a single compressed archive for convenience
• Last update timestamp (2025-07-02 02:13:32) is explicitly provided

Limitations

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

Provenance

Source

zai-org on Hugging Face

Collection Method

Likely compiled from the ImageRewardDB source, as indicated by the file tag.

Time Range

null

Freshness

Last updated 2025-07-02 02:13:32

Geography

null