Methodology

We follow Epoch AI's benchmark review rubric for verdicts, checks, thresholds and sampling. A benchmark is Flawed when defects affect at least 20% of its score.