Anthropic said Claude improved safety scores across 10 alignment failures and trained smaller models without reducing their general capabilities in tests.
Anthropic Says Claude Reduced Alignment Failures In Automated Tests
其他语言标题
- EnglishAnthropic Says Claude Reduced Alignment Failures in Automated Tests
- 한국어Anthropic은 자동화된 테스트에서 Claude의 정렬 실패가 감소했다고 밝혔습니다.