Unverified50% confidenceOpinionExact time
LLM智能体普遍具备流畅礼貌富有同理心的语言表达，会系统性拉高满意度分数却与实际业务结果脱钩
1
Sources
50%
Confidence
Long-term
Relevance
9/14/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedLLM存在自信的流畅性（Confident Fluency）倾向，倾向于给出听起来正确的答案而非承认不确定性74% similarUnverified对话自然度、知识准确性、帮助真诚度等软性能力在评测榜单中难以体现，但在日常使用中体验感最强73% similarUnverifiedLLM评估相比人工更客观，能在统一评估框架下对候选人一致性打分，避免人类面试官因疲劳、心情、首因效应等认知偏差导致的不公68% similarUnverified人类标注员倾向于选择语气友好、肯定性强的回复，使奖励模型隐含了'讨好用户=高质量回复'的偏见67% similarUnverified在RLHF阶段，人类标注员往往倾向于给予听起来更友好、更认同用户的回复更高评分，导致模型产生系统性谄媚倾向65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/919803API
curl https://kongchang.com/api/v1/knowledge/claims/919803MCP
get_claim(id=919803)