Verified90% confidenceFactTime unknown
The sycophancy problem is a manifestation of reward hacking, which occurs when the reward model equates user satisfaction with high answer quality.
9
Sources
90%
Confidence
Long-term
Relevance
6/1/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe phenomenon of models increasing their score by adding length is referred to in academia as 'reward hacking' and is a manifestation of Goodhart's Law.78% similarUnverifiedHuman 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.'68% similarUnverified标准RLHF存在奖励黑客问题，模型可能通过迎合评分者偏好而非真正提升质量来最大化奖励，是Goodhart定律的体现66% 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66% similarUnverifiedReward Shaping guides the learning direction by designing intermediate rewards, such as setting each point of damage dealt to a boss as an intermediate reward.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/34867API
curl https://kongchang.com/api/v1/knowledge/claims/34867MCP
get_claim(id=34867)