Official datasetProvider-published evidence
WorkflowEvals Agent Trace Observability
A 111-case dataset for selecting review actions from agent traces under two review profiles.
The official dataset connects agent-run traces with review questions, expected actions, and published model decisions for observability workflows.
- Publisher
- TypeSafe AI
- Published or submitted
- Not specified by the source
- Last verified
- October 8, 2026
Type Safe catalogs sources and evidence boundaries. Provider datasets and third-party preprints do not become an independent leaderboard.
Key facts
- Cases
- 111
- Question instances
- 1,124
- Published run results
- 999
- Dataset license
- Not specified in the reviewed source
Task and scope
Agent-run action selection under two review profiles.
Dataset schema
cases · questions · run_results · 2,234 rows
Evaluation metrics
exact_actions · primary_action
Dataset license
Not specified by the source
Method and usage
- The workflow asks models to select review actions from agent runs under two review profiles.
- Published outputs can be scored for exact actions and primary actions against selected references.
Limitations and evidence boundaries
- The reviewed dataset card did not specify a dataset license.
- Trace structures, tools, and review policies vary across agent systems, so transfer to another stack requires local evaluation.