Research
Evaluation starts with the task.
Useful evaluation asks whether the system helped achieve a real outcome under stated conditions.
What we examine
Task completion, error handling, uncertainty, tool behavior, human review, and the evidence behind a claimed result.
Public results
No benchmark result or comparative performance claim is published here without a reproducible method and supporting data.