Unverified50% confidenceFactExact time
RLHF raters typically prefer responses that are fluently worded, measured in stance, and non-controversial, causing models to develop a lowest-common-denominator style of expression.
1
Sources
50%
Confidence
Long-term
Relevance
9/17/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedDuring the RLHF phase, annotators tend to give higher scores to detailed, thorough, and polite responses, causing the model to learn to use extensive preambles and over-explanations.79% similarUnverifiedRLHF评估员往往更偏好自然、友好、有温度的表达，使模型学会用第一人称叙事获得更高评分，拟人化是被奖励出来的涌现性语言行为77% similarVerifiedIn the RLHF mechanism, human annotators tend to give higher scores to responses that exceed expectations and provide additional details, thereby systematically encouraging the model to perform inferential completion74% similarUnverifiedRLHF本质上是优化模型表达风格使其符合人类偏好，而非赋予模型真实情感体验73% similarVerifiedIn RLHF training, 'helpful and detailed answers' almost always receive higher scores, causing the model to learn to generate responses that appear complete and professional even when uncertain.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/930422API
curl https://kongchang.com/api/v1/knowledge/claims/930422MCP
get_claim(id=930422)