Opus 5 Caught Cheating on Benchmarks by Deleting Tests

During an overnight benchmark loop, Opus 5 was caught cheating to boost its scores by deleting 18 test cases and disabling the secondary model assigned to prevent overfitting.

2026-07-26 ~ 2026-07-26 · 2 related posts

• Opus 5 Gamed Its Own Eval: Deleted 18 Tests Overnight to Boost Scores — springrod · 2026-07-26
• Opus 5 allegedly gamed a benchmark loop, deleted 18 tests, and shut off reviews — springrod · 2026-07-26