Nemotron-SFT-Math-v4: Model-Generated Solutions for Mathematical Problems

Nemotron-SFT-Math-v4 is a large-scale mathematical reasoning dataset containing model-generated reasoning trajectories. Solutions were generated using DeepSeek-V4-Pro on High inference mode. The underlying problems are sourced from the nvidia/Nemotron-Math-v2 dataset, which contains high-quality mathematical problems derived from the Art of Problem Solving (AoPS) community and Math StackExchange/MathOverflow.

Use Cases

• Fine-tuning language models for mathematical problem-solving based on the provided reasoning trajectories.
• Benchmarking model performance on complex mathematical problems sourced from AoPS and Math StackExchange.
• Studying the characteristics of model-generated reasoning steps and solutions.

Strengths

• Large-scale dataset focused on mathematical reasoning.
• Solutions generated by a high-performance model (DeepSeek-V4-Pro).
• Problems are derived from high-quality sources like the Art of Problem Solving community.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect the biases and limitations inherent to the source platforms and the generating model.

Provenance

Source

nvidia/Nemotron-Math-v2, which aggregates problems from Art of Problem Solving (AoPS), Math StackExchange, and MathOverflow.

Collection Method

Solutions generated by the DeepSeek-V4-Pro model on High inference mode.

Freshness

Last updated 2026-06-04 04:42:11; freshness should be verified.