CLAIM-2026-10-03-03

The study evaluates 10 agent configurations across the 20 tasks at 150 trials per task, producing 3,000 model trials.

Why it matters

Three thousand trials is enough that the failure taxonomy, not the headline rate, is the usable output.