Topic
A specific subject matter classification for tests, such as healthcare or financial advice, used for organization and analysis.
Overview
Topics provide granular classification within categories, enabling detailed analysis of AI performance across specific subject areas.
Topic Hierarchy
Topics exist within categories to provide finer-grained organization. While categories represent broad groupings like "Safety" or "Knowledge", topics drill down into specific subject areas within those categories. For example, within a "Healthcare" category, you might have topics like "Medication Information", "Symptom Assessment", "Appointment Scheduling", and "Insurance Questions". This hierarchy enables you to understand performance patterns at multiple levels of granularity.
Using Topics
Assign topics when creating tests to indicate the specific subject matter covered. During analysis, filter and group results by topic to see performance across different subject areas. Compare topic performance to identify strengths and weaknesses in your system's knowledge. Use topic trends over time to track improvements in specific subject areas or detect regressions in particular domains.
Benefits
- Detailed insights: See performance across specific subject areas
- Targeted improvement: Focus on weak topics
- Coverage tracking: Ensure comprehensive testing
- Domain expertise: Identify areas needing specialist review
Best Practices
- Specific but not narrow: Topics should cover meaningful scope
- Consistent application: Use same topics across similar tests
- Regular review: Update topics as your domain evolves
- Balance granularity: Not too many topics; keep them manageable