Read scores and inspect results
Understand the measurement before deciding what it supports.
Scoring compares predictions with compatible Ground Truth using the selected Test plan and settings. Open Scoring in the Run to inspect the saved result and supporting examples.
Confirm the basis
Check the Dataset and label versions, task, inference source, sample coverage, and selected metrics. Predictions, Ground Truth, and criteria have distinct roles; changing one can change the meaning of the result.
Interpret the measurement
Use the metric reference to understand the task-specific score. Examine per-class or sample details where supplied and investigate examples behind an aggregate value. Compare conditions only with their source and coverage visible.
Read the criterion separately
A measured score states what happened under the recorded conditions. An acceptance criterion states what the work was intended to meet. Keep both visible in the interpretation.
Re-score or retain the result
Re-scoring preserves earlier results and the Test plan association. For an Assessment requirement, link the selected result and record the human conclusion. For a handoff, review the report and its limitations.