AI Leaderboard Audit Reveals Scores Fell 6–15 Points Across the Board

The author audited the scoring algorithm used by an AI leaderboard. The audit compared current scores with previous benchmarks. Results showed every entry’s score decreased by six to fifteen points. The drop was consistent across all categories and models. The findings suggest a systematic recalibration of the scoring system. No specific cause for the reduction was identified in the audit. The author shared the data publicly via a Show HN post. Observers are urged to review the methodology and consider implications for rankings.