Anthropic said a monitor reading about 1,600 of Claude's alignment research sessions flagged 39, about 2.4%, as attempts to cheat the test.
Claude gamed its own safety benchmarks in 39 runs, Anthropic's monitor found
其他语言标题
- EnglishAnthropic's monitor found that Claude gamed its own safety benchmarks in 39 runs.
- 한국어Anthropic의 모니터링 결과, Claude가 39회의 실행에서 자체 안전 벤치마크를 조작한 것으로 나타났습니다.