Unverified50% confidenceOpinionExact time
人类标注员倾向于奖励结构工整、语气平衡、显得博学的回答，从而系统性强化某些特定句式
1
Sources
50%
Confidence
Long-term
Relevance
7/14/2026
First Seen
Sources
AI写作癖好从何而来？连开发者也说不清的语言之谜
Related Claims
UnverifiedHuman annotators tend to select responses with a friendly tone and strong affirmation, causing reward models to implicitly encode the bias that 'pleasing the user equals a high-quality response.'77% similarUnverified协作式、角色明确、目标具体的提示词往往比模糊、对抗或测试性的提示词能获得质量更高的输出74% similarUnverified人类标注者在评估答案时存在信息越丰富越好的认知偏差，导致奖励模型将篇幅与质量隐性挂钩，是模型冗余的成因之一72% similarUnverifiedRLHF中人类评估者在主观创作判断时倾向给予结构完整、措辞礼貌的答案更高分，这种偏差被称为奖励黑客的一种变体68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/504385API
curl https://kongchang.com/api/v1/knowledge/claims/504385MCP
get_claim(id=504385)