Instruction Following Prompts for Language Model Evaluation

Nemotron-RL-instruction_following combines prompts from the WildChat-1M dataset with verifiable instructions from the Open-Instruct code base. Created by NVIDIA, this dataset is designed for training and evaluating models on objective instruction adherence. It was last updated in January 2026.

Use Cases

• Fine-tuning language models to follow verifiable instructions, such as the 'response under 200 words' constraint.
• Benchmarking model performance on objective adherence metrics using prompts from the WildChat-1M source.
• Training reward models for RLHF by assessing compliance with specific, measurable instructions in the dataset.

Strengths

• Derived from the established WildChat-1M dataset, providing a foundation of diverse user prompts.
• Instructions are designed for easy verification, enabling clear evaluation of model adherence.

Limitations

• Specific dataset size, row count, and column details are unknown.
• Potential bias or limitations inherited from the source WildChat-1M dataset are not detailed.

Provenance

Source

Combines prompts from WildChat-1M and instructions from the Open-Instruct code base.

Collection Method

Created by combining and curating data from the two specified sources.

Freshness

Last updated in January 2026.