HelpSteer is a dataset released by NVIDIA in 2023 for aligning large language models, designed to improve the helpfulness, correctness, and coherence of model responses, while also supporting adjustments to the complexity and verbosity of replies. The related research paper is "HelpSteer: Multi-attribute Helpfulness Dataset for SteerLM".

The dataset contains 37,120 samples, covering various task types such as rewriting, summarization, classification, extraction, and question answering. Each sample includes a prompt, a model response, and five attribute scores annotated by humans—helpfulness, correctness, coherence, complexity, and verbosity—each ranging from 0 to 4. The data was annotated by Scale AI and is primarily used for large language model training, evaluation, and alignment research.

Dataset Composition

The dataset is divided into two parts: training set and validation set:

• Training set (train): 35,331 samples.
• Validation set (validation): 1,789 samples.

Each sample mainly contains the following fields:

• prompt: The input prompt.
• response: The response generated by the large language model.
• helpfulness: Helpfulness score.
• correctness: Correctness score.
• coherence: Coherence score.
• complexity: Complexity score.
• verbosity: Verbosity score.

Citation

@misc{wang2023helpsteer,
      title={HelpSteer: Multi-attribute Helpfulness Dataset for SteerLM}, 
      author={Zhilin Wang and Yi Dong and Jiaqi Zeng and Virginia Adams and Makesh Narsimhan Sreedhar and Daniel Egert and Olivier Delalleau and Jane Polak Scowcroft and Neel Kant and Aidan Swope and Oleksii Kuchaiev},
      year={2023},
      eprint={2311.09528},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}

@misc{dong2023steerlm,
      title={SteerLM: Attribute Conditioned SFT as an (User-Steerable) Alternative to RLHF}, 
      author={Yi Dong and Zhilin Wang and Makesh Narsimhan Sreedhar and Xianchao Wu and Oleksii Kuchaiev},
      year={2023},
      eprint={2310.05344},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}