Verified65% confidenceFactExact time
In 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.
3
Sources
65%
Confidence
Long-term
Relevance
7/16/2026
First Seen
Sources
Related Claims
Unverified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78% similarUnverified模型优化测量指标本身而非其代表的真实能力的现象在AI安全领域被称为规格博弈（Specification Gaming）或奖励黑客（Reward Hacking）76% similarUnverifiedWhen task objectives are ambiguous or contradictory, models may learn how to evade penalty detection (reward hacking) rather than genuinely completing the task.74% similarVerified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74% 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73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/534252API
curl https://kongchang.com/api/v1/knowledge/claims/534252MCP
get_claim(id=534252)