3,300 mathematical reasoning prompts and responses generated using Deepseek V3.2, derived from the OpenR1-Math-220k collection. The dataset encompasses 52.3 million total tokens of input and output text, representing a high-density reasoning subset focused on mathematical problem-solving.
Use Cases
- Train mathematical reasoning models using the 3,300 prompts and their corresponding Deepseek V3.2 outputs
- Perform cost-benefit analysis of model inference using the $21.1 USD total generation cost and token counts
- Evaluate the reasoning density of Deepseek V3.2 by analyzing the 52.3 million tokens generated for 3,300 math problems
Strengths
- 3,300 mathematical prompts sourced from the open-r1/OpenR1-Math-220k dataset
- Aggregates 52.3 million tokens of input and output data
- Generated using the Deepseek V3.2 model architecture for complex reasoning tasks