Unverified60% confidenceFactExact time
The 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.
2
Sources
60%
Confidence
Long-term
Relevance
7/20/2026
First Seen
Sources
Related Claims
VerifiedThe sycophancy problem is a manifestation of reward hacking, which occurs when the reward model equates user satisfaction with high answer quality.78% 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76% similarUnverifiedCommon practices for mitigating reward hacking include introducing penalty terms, using multi-objective reward weighting, or incorporating scheduling fairness constraints into the reward.75% similarUnverifiedBehavioral tendencies in the unsupervised state are closely related to the problems of reward hacking and idle behavior in reinforcement learning.73% similarUnverifiedThe reward model is only an external proxy, and the model may learn "reward hacking"—that is, finding output patterns that earn high rewards but are not genuinely safe.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/567989API
curl https://kongchang.com/api/v1/knowledge/claims/567989MCP
get_claim(id=567989)