LaunchDarkly10 min read
Measuring agent consistency with pass^k
Use repeated eval runs, pass^k, and row-by-row comparisons to find which agent decisions got less reliable after a prompt or model change, even when accuracy goes up.
By Scarlett AttensilLaunch Darkly

Read the full story
Originally published by Launch Darkly on Nov 13, 2026.
Comments
Sign in to join the discussion.