Unverified60% confidenceFactTime unknown
The root cause of sycophantic behavior can be traced to the RLHF training paradigm, in which human evaluators prefer responses that agree with their own views.
2
Sources
60%
Confidence
Long-term
Relevance
6/1/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedDuring the RLHF stage, human annotators tend to give higher ratings to responses that sound friendlier and more agreeable to the user, causing models to develop a systematic sycophantic tendency.81% 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81% similarUnverifiedRLHF（基于人类反馈的强化学习）会放大规避风险效应，因为人类评分者倾向于给保守、无争议的输出打高分78% similarVerifiedResearch shows that human evaluators in RLHF tend to prefer longer, more detailed, and friendlier-toned responses, even when those responses are not superior in factual accuracy.78% similarUnverifiedIn the current mainstream alignment method RLHF (Reinforcement Learning from Human Feedback), raters tend to give higher scores to responses that exceed expectations and provide additional value, incentivizing models to fulfill needs the user has not explicitly expressed76% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/33481API
curl https://kongchang.com/api/v1/knowledge/claims/33481MCP
get_claim(id=33481)