Nemotron RLHF Training Prompts and Metadata

45,882 samples comprise this Reinforcement Learning from Human Feedback training dataset. NVIDIA created it for language model alignment, with the dataset last updated in December 2025.

Use Cases

• Training reward models using prompts and their associated category information.
• Fine-tuning language models via RLHF on the provided prompt data.
• Analyzing prompt distribution and characteristics across different data sources.
• Benchmarking alignment techniques against a set of 45,882 RLHF training samples.

Strengths

• Contains 45,882 samples for RLHF training.
• Includes prompts, data sources, and category information per sample.
• Designed and released by NVIDIA, a major AI research organization.

Limitations

• Specific column names and data structure are not publicly documented.
• Sample data is unavailable for inspection prior to download.
• The dataset's composition and potential biases are not described.

Provenance

Source

NVIDIA

Freshness

Last updated December 2025.