Unverified75% confidenceFactTime unknown
Anthropic在设计Claude人格时一直在试图平衡友善和诚实这两个目标
1
Sources
75%
Confidence
Long-term
Relevance
6/1/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAnthropic在模型对齐研究中将诚实性（Honesty）列为核心价值观66% similarPartially VerifiedAnthropic uses an automated classifier (the LLM-as-a-Judge method) to evaluate Claude's degree of sycophancy across four dimensions: willingness to contradict, firmness of position, appropriateness of praise, and frankness.62% similarUnverifiedAnthropic proactively disclosed Claude's sycophancy rate data rather than concealing information that is unfavorable to the company.60% similarUnverifiedAnthropic's Claude series has long enjoyed a real-world product reputation for long-document understanding and instruction-following that surpasses some competitors with higher benchmark scores.59% similarUnverifiedAnthropic的研究将对齐细分为行为对齐、意图对齐和动机对齐三个层次58% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/35106API
curl https://kongchang.com/api/v1/knowledge/claims/35106MCP
get_claim(id=35106)