Multi-Cultural Single/Multi-Agent Images: 9,000 Generated Images for Cultural Interaction

9,000 images generated for the novel task of multicultural text-to-image generation. The dataset forms the empirical foundation of the MosAIG framework introduced in the paper 'Multi-Agent Multimodal Models for Multicultural Text to Image Generation.' It was created by AIM-SCU and last updated on HuggingFace in April 2026.

Use Cases

• Benchmarking text-to-image models for cultural fairness based on depictions of cross-cultural interactions.
• Training multimodal AI models to generate images from prompts involving agents from one culture and landmarks from another.
• Researching bias and representation in generative AI based on the described cross-cultural scenarios.
• Developing evaluation metrics for multicultural image generation based on the described framework.

Strengths

• Contains 9,000 images, providing a substantial corpus for evaluation.
• Specifically designed for the novel task of multicultural text-to-image generation, addressing a defined research gap.
• Serves as the empirical foundation for the published MosAIG framework.

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

AIM-SCU via HuggingFace.

Collection Method

Images were generated, likely using AI models, for the described research task.

Freshness

Last updated 2026-04-20 23:45:34; freshness should be verified.