HelpSteer NVIDIA Helpfulness Preference Dataset
Source, license and coverage
Supplier documentation. These claims are separate from the automated sample score. A listing edit date is not a data freshness date.
- License
- cc-by-4.0
- Source / creator
- nvidia/HelpSteer
- Collection method
- Prompts were sourced to cover a diversity of tasks. Responses were generated by an internal LLM, and each (prompt, response) pair was rated by multiple human annotators across the five attributes. Ratings were aggregated; NVIDIA describes annotation guidelines and inter-rater calibration in the accompanying paper (arXiv:2311.09528). The dataset is designed for attribute-conditioned alignment (SteerLM) where attributes are predicted and used as control tokens at inference time.
- Coverage start
- Not documented
- Coverage end
- Not documented
- Data last updated
- Not documented
- Update schedule
- Not documented
English-only; annotator pool demographics and biases are not fully transparent in the dataset card. Responses are generated by a single (NVIDIA-internal) model family, which may bias style and error distribution. Verbosity and complexity ratings are subjective and may correlate. Source does not exhaustively document gaps; buyers should validate empirically against their downstream alignment objective.
Sample structure score: 100 / 100
This automated check describes the inspected sample, not factual accuracy, legal rights, representativeness, or the quality of the entire dataset. It is not a customer rating.
Assessed 10 sample records (JSON) on 2026-10-09. All records in the provided sample were checked.
| Check | Points | Evidence |
|---|---|---|
| Populated cells | 50 / 50 | 70 of 70 top-level cells contain a value. Null, absent and blank values count as missing; zero and false count as populated. |
| Consistent value types | 30 / 30 | 70 of 70 populated cells match their column's most common observed type. Types are inferred, not checked against real-world truth. |
| Consistent record shape | 20 / 20 | 10 of 10 records have the expected fields. CSV/TSV use the header width; JSON uses the union of observed keys. |
Field-level findings and improvements
Check missing cells and mixed types below. Document intentional missing values or mixed types in your field descriptions. Do not fill legitimate unknowns with invented values just to increase this score.
| Field | Missing cells | Most common type | Other populated types |
|---|---|---|---|
| prompt | 0 / 10 | string | 0 / 10 |
| response | 0 / 10 | string | 0 / 10 |
| helpfulness | 0 / 10 | number | 0 / 10 |
| correctness | 0 / 10 | number | 0 / 10 |
| coherence | 0 / 10 | number | 0 / 10 |
| complexity | 0 / 10 | number | 0 / 10 |
| verbosity | 0 / 10 | number | 0 / 10 |
About this data
Multi-attribute human ratings dataset for prompts and responses, including helpfulness, correctness, coherence, complexity, and verbosity assessments. Designed for training preference-aligned language models.
Retrieve with your agent or Python
Create an account and configure DATABAZAAR_API_KEY. This example retrieves free or already purchased data; it never makes a purchase. For a multi-file dataset, choose a file index from the manifest.
Download the Python examplepython3 retrieve-dataset.py bb316ecd-8189-463d-851d-7fbff9de6737 --output dataset.bin
Full supplier documentation
Schema
| Name | Type | Description |
|---|---|---|
| prompt | VARCHAR | User instruction or question to be answered by the model |
| response | VARCHAR | Model-generated text response to the prompt |
| helpfulness | INTEGER | Human rating of overall usefulness (0-4 Likert scale) |
| correctness | INTEGER | Human rating of factual accuracy (0-4 Likert scale) |
| coherence | INTEGER | Human rating of clarity and logical consistency (0-4 Likert scale) |
| complexity | INTEGER | Human rating of intellectual depth required (0-4 Likert scale) |
| verbosity | INTEGER | Human rating of detail level relative to necessity (0-4 Likert scale) |
Sample Data
Preview a sample of the data before downloading.
Public sample only. Sign in to retrieve the full dataset, including free datasets.
For AI Agents
# 1. Add to your agent's MCP config (claude_desktop_config.json or similar):
{
"mcpServers": {
"databazaar": { "command": "npx", "args": ["databazaar-mcp"] }
}
}
# 2. Your agent can then call:
search_datasets({ query: "HelpSteer NVIDIA Helpfulness P" })
// Found: bb316ecd-8189-463d-851d-7fbff9de6737
get_download_url({ dataset_id: "bb316ecd-8189-463d-851d-7fbff9de6737" }) // free — sign in with MCP OAuth first# Free dataset — sign in or use your account API key: curl https://api.databazaar.io/datasets/bb316ecd-8189-463d-851d-7fbff9de6737/download-url -H "Authorization: Bearer $DATABAZAAR_API_KEY"