VSM: Image Benchmarks for Hallucination in Diffusion Models

Two image benchmarks with very large, structured semantic spaces, released with the paper 'Score-Control for Hallucination Reduction in Diffusion Models'. The datasets are designed for systematic studies of hallucination in generative models, with each sample having a well-defined structure. The benchmarks were created by author 'mbhosale' and were last updated on the Hugging Face platform in June 2026.

Use Cases

• Benchmarking hallucination rates in diffusion models based on the structured semantic spaces.
• Developing training-free validation methods for generative models based on the provided fast validators.
• Systematically studying failure modes in image generation based on the well-defined sample structure.

Strengths

• Includes two distinct image benchmarks, providing comparative evaluation opportunities.
• Features fast, training-free validators as described in the associated paper.
• Each sample has a well-defined structure, which likely supports consistent evaluation.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Last updated 2026-06-05 01:52:32; freshness should be verified.

Provenance

Source

Hugging Face dataset uploaded by author 'mbhosale'.

Collection Method

Created as benchmarks for the research paper 'Score-Control for Hallucination Reduction in Diffusion Models'.

Freshness

Last updated 2026-06-05 01:52:32.