README.md

metadata

license: odc-by
language:
  - en
configs:
  - config_name: default
    data_files:
      - split: train
        path: ifrl_final_release.parquet

Dataset Description:

The Nemotron-Cascade-RL-IF-RL dataset is designed for Instruction-Following Reinforcement Learning (IF-RL). It contains prompts and associated metadata for improving language model's instruction following capability.

This dataset is ready for commercial use (with attribution).

The dataset contains the following subset:

Training Data

This data contains 108,938 samples used for IF-RL training. It includes prompts, data sources, and instruction-following meta annotations required for the rule verifier. The data sources are the following:

• Filtered and pre-processed Llama-Nemotron-Post-Training-Dataset
• Augmented instruction-following data using the prompts from LMSYS-Chat-1M.

Dataset Owner(s):

NVIDIA Corporation

Dataset Creation Date:

Created on: Dec 15, 2025 Last Modified on: Dec 15, 2025

License/Terms of Use:

The dataset is governed by the ODC-BY-1.0.

Intended Usage:

This dataset is intended to be used by the community to train language models to have instruction-following capability. The data may be freely used to train and evaluate.

Dataset Characterization

Data Collection Method
Hybrid: Human, Synthetic, Automated

Labeling Method
Hybrid: Human, Synthetic, Automated

Dataset Format

Modality: Text
Format: Parquet
Structure: Text + Metadata

Columns:

• prompt: The input prompt for the model (Chat format)
• instruction_id_list: annotation required for instruction following rule verifier
• kwargs: annotation required for instruction following rule verifier
• index: Identifier

Dataset Quantification

Subset	Samples
train	108,938
Total	108,938

Total Disk Size: 26MB

Ethical Considerations:

NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. When downloaded or used in accordance with our terms of service, developers should work with their internal developer teams to ensure this dataset meets requirements for the relevant industry and use case and addresses unforeseen product misuse.
Please report quality, risk, security vulnerabilities or NVIDIA AI Concerns here