Unverified50% confidenceFactExact time
Anthropic研究人员在HHH框架论文中承认Helpful、Harmless、Honest三者之间存在内在张力，过分强调Harmless会损害Helpful
1
Sources
50%
Confidence
Long-term
Relevance
7/13/2026
First Seen
Sources
从热爱到失望：Claude新模型体验下滑的深层原因
hackernewshackernews7/11/2026
Related Claims
VerifiedAnthropic将AI对齐的核心原则称为HHH原则（Helpful, Honest, Harmless）72% similarUnverified对齐的目标通常被概括为3H原则：有用（Helpful）、诚实（Honest）和无害（Harmless）66% similarUnverified主流AI安全训练遵循HHH原则——Helpful（有帮助）、Harmless（无害）、Honest（诚实），这三个目标在情感敏感领域会产生严重冲突62% similarUnverifiedAnthropic的模型规范（Model Spec）要求Claude同时具备有帮助（helpful）、诚实（honest）和无害（harmless）三个属性，即3H原则62% similarUnverified多智能体的分而治之理念由 Anthropic 提出57% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/500389API
curl https://kongchang.com/api/v1/knowledge/claims/500389MCP
get_claim(id=500389)