RL Mixed Dataset: Multimodal Math and Geometry Problems for Reinforcement Learning

RL Mixed Dataset is a multimodal collection created by hiyouga by merging the geometry3k and math12k datasets. The dataset was last updated on June 11, 2025. It likely contains image-text pairs for mathematical and geometric problems intended for reinforcement learning research.

Use Cases

• Training multimodal AI agents on geometry problems based on the described image-text pairs.
• Benchmarking reinforcement learning models for mathematical reasoning tasks.
• Developing visual question-answering systems for educational content.
• Creating synthetic training data for models that combine computer vision and natural language processing.

Strengths

• Combines two established datasets (geometry3k and math12k), suggesting a merged scale.
• Dataset was updated on June 11, 2025, indicating recent maintenance.

Limitations

• Row count, column definitions, and file formats are unknown, limiting suitability assessment.
• Description metadata is limited; actual data quality and structure require manual inspection after download.
• The license is unspecified, which may restrict usage.

Provenance

Source

hiyouga on Hugging Face, merging datasets from the same author.

Collection Method

Merged from hiyouga/geometry3k and hiyouga/math12k using a custom script.

Freshness

Last updated 2025-06-11 12:59:20