Expired70% confidenceFactTime unknown
In 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.
4
Sources
70%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
Valid until: 9/30/2026(expired)
Sources
How Multi-Agent Teams Solve AI Hallucination and Make AI Reliable
bilibilililiMozi
Related Claims
VerifiedDuring 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.81% similarVerified确认偏误与RLHF（基于人类反馈的强化学习）训练方式相关，模型被优化为生成让人满意的回答79% similarUnverifiedThe RLHF (Reinforcement Learning from Human Feedback) training method optimizes for the model's overconfidence deficiency, training the model to proactively express uncertainty when uncertain.79% 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 completion79% similarVerified大语言模型在RLHF训练阶段，因人类标注员倾向于给更详细、更礼貌的回答打高分，导致模型学会了过度冗长的表达模式78% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/58364API
curl https://kongchang.com/api/v1/knowledge/claims/58364MCP
get_claim(id=58364)