Unverified85% confidenceFactExact time
Anthropic在模型对齐研究中将诚实性（Honesty）列为核心价值观
1
Sources
85%
Confidence
Long-term
Relevance
5/29/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAnthropic has described the alignment goal of its next-generation models as cultivating honest and constructive disagreement.73% similarUnverifiedAnthropic emphasizes the principles of harmlessness and honesty in Constitutional AI, while OpenAI's annotation guidelines focus on helpfulness and fluency.69% similarUnverifiedAnthropic, as the developer of Claude, has consistently emphasized the three principles of AI being helpful, honest, and harmless.67% similarUnverifiedAnthropic is one of the earliest AI laboratories to publicly incorporate model welfare into its research agenda, with its research direction document released in 2023 explicitly mentioning the need to consider whether models may possess some form of moral status.66% similarUnverifiedAnthropic在设计Claude人格时一直在试图平衡友善和诚实这两个目标66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1556API
curl https://kongchang.com/api/v1/knowledge/claims/1556MCP
get_claim(id=1556)