VP-Bench is the official evaluation benchmark for the FlowInOne model, which is described as a vision-centric image-in, image-out generation model. The benchmark is curated to assess instruction faithfulness, spatial precision, visual realism, and content. It was created by CSU-JPG and last updated on Hugging Face in April 2026.
Use Cases
- Benchmarking instruction faithfulness in image generation models based on the described evaluation criteria
- Evaluating spatial precision in generated images based on the described evaluation criteria
- Assessing visual realism of generated content based on the described evaluation criteria
- Comparing content quality of image-in, image-out models based on the described evaluation criteria
Strengths
- Officially curated benchmark for the FlowInOne model
- Designed to assess multiple specific criteria: instruction faithfulness, spatial precision, visual realism, and content
Limitations
- Column-level documentation is absent; field semantics must be inferred after download
- Row count is unknown, which may limit suitability assessment
- Description metadata is limited; actual data quality requires manual inspection after download
Provenance
- Source
- CSU-JPG
- Collection Method
- Rigorously curated benchmark, likely for academic evaluation
- Time Range
- null
- Freshness
- Last updated 2026-04-09 02:18:45
- Geography
- null