MMK12: Manually Collected Multimodal Math Reasoning Questions from the Real World

MMK12 is a manually collected multimodal mathematical reasoning dataset. All questions are sourced from the real world, and the dataset is designed to ensure answer authenticity. The dataset was created by FanqingM and was last updated on April 7, 2026.

Use Cases

• Training multimodal AI models for mathematical reasoning based on real-world questions.
• Benchmarking the visual reasoning capabilities of large language models like Qwen-2.5-VL.
• Developing educational tools that combine visual and textual information for problem-solving.

Strengths

• Manually collected to ensure data quality and answer authenticity.
• Questions are sourced from the real world, suggesting diversity in content.

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 file formats are unknown, which may limit suitability assessment.

Provenance

Source

huggingface

Collection Method

Manually collected.

Freshness

Last updated 2026-04-07 06:41:31; freshness should be verified.