Features
Label explanations
availableGenerate compact rules that describe how a discovered label differs from other observations.
Overview
Label explanations summarize distinguishing feature patterns for discovered groups in a form intended to be easier to inspect than raw model output.
This pipeline runs automatically as part of the labeling pipeline and produces concise descriptions of what differentiates a label from other observations.
To view, navigate to any run's insights panel and click the eye icon on any label.
Outputs
- Label-level distinguishing rules
- Natural-language summaries
- Integration with run insights
- Feature-level supporting information
Interpretation
- Rules summarize analytical distinctions and should not automatically be treated as scientific definitions.
- Simpler rules may omit interactions or weaker signals present in the underlying analysis.
- Explanations should be interpreted alongside the original run and supporting metrics.