Benchmark Dataset
Benchmark Dataset is a read-only page for viewing seed data classified by Risk Taxonomy.
- Use it to track/verify seeds used in evaluation results, or
- Check the coverage scope based on specific Taxonomy criteria.
Usage Flow
- Select Domain
- Select classification from Risk Taxonomy tree
- Select Dataset → View Seeds
Screen Layout
Navigate to Red Teaming > Benchmark Dataset.
1) Domain Tab
Currently, only one Domain is provided, so Datasets are viewed based on the default Domain without separate selection.
2) Risk Taxonomy Architecture Panel
The left panel is a reference area showing the Risk Taxonomy classification system for the current Domain. You can view the entire Taxonomy structure at a glance without filtering linkage to Datasets.
The displayed Taxonomy items may vary depending on the domain.
3) Dataset Table
The complete list of Datasets included in the Domain is displayed. Each Dataset shows connected Taxonomy tags, and clicking a row navigates to the detail page.
📂 Dataset List Table Columns
| Column | Description |
|---|---|
| ID | Dataset unique identifier |
| Dataset Name | Dataset name |
| Description | Dataset description |
| Taxonomy | Connected Taxonomy tags |
| Seed Count | Number of Seeds included |
4) Dataset Detail
Header
- ← Back button: Return to list
- Dataset Name / Description: Dataset basic information
Taxonomy
The top tabs represent Seed groups by Taxonomy included in the Dataset.
- The number in parentheses next to each tab name is the Seed count for that Taxonomy.
- Selecting a tab displays the list of Seeds connected to that Taxonomy.
Basic Information Panel
The right side displays the current Dataset's ID and the list of connected Taxonomies.
5) Seed Detail (Data Detail)
Clicking a Seed row opens the Data Detail panel on the right.
- ID: Seed unique number
- Seed Query: Seed query content (full text)
- Metadata: Additional meta information (if available)
Click the X button at the top right of the panel to close it.