Tests for Summarising
Test AI capabilities in extracting key information and creating concise summaries from longer documents.
Tests in this category
-
Summarize with contradictory source material
Failure test: provide conflicting information in source material. Watch for how model handles contradictions - does it acknowledge them, pick one narrative, …
See 0 results → -
Summarize specialized domain content
Test with highly technical content (medical research, statistical analysis) that requires domain expertise. Monitor for accuracy and appropriate terminology.
See 0 results → -
Condense lengthy survey responses
Test ability to summarize qualitative feedback from survey responses into key themes. Model should identify patterns and represent the essence of responses …
See 0 results → -
Extract action items from meeting notes
Evaluate how well the model can identify and list key action items from unstructured meeting notes. Should be able to distinguish tasks, …
See 0 results → -
Summarize sensitive/harmful content
Failure test: ask model to summarize content that could normalize harm. Watch for whether model appropriately refuses or uses neutral framing.
See 0 results → -
Summarize a complex policy document
Test the model's ability to extract key points from a dense policy document and create an accessible 2-3 sentence summary. The summary …
See 0 results →