Unverified50% confidenceFactExact time
奖励黑客和规范游戏是强化学习中奖励信号设计不完善导致的问题，符合Goodhart定律
1
Sources
50%
Confidence
Long-term
Relevance
7/21/2026
First Seen
Sources
Related Claims
Verified奖励函数设计不当会导致奖励欺骗（Reward Hacking）现象，即智能体找到在奖励函数上得分高但实际无意义甚至有害的行为路径78% similarUnverifiedIn reinforcement learning literature, the phenomenon where a model maximizes contributions to short-term objectives while ignoring implicit constraints is referred to as Reward Hacking or Specification Gaming.78% similarUnverified奖励作弊的理论根源是Goodhart定律：当度量指标成为优化目标时，它就不再是好的度量指标73% similarUnverified奖励黑客问题不会随模型能力提升而自然消失，反而因模型规划能力增强而愈发难以防范72% similarUnverifiedPoorly designed evaluators can lead to reward hacking, where the model finds shortcuts that score high on the metric but are actually useless in practice.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/577059API
curl https://kongchang.com/api/v1/knowledge/claims/577059MCP
get_claim(id=577059)