Unverified50% confidenceFactExact time
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）
1
Sources
50%
Confidence
Long-term
Relevance
7/18/2026
First Seen
Sources
Related Claims
UnverifiedWhen AI optimizes toward passing tests, it may pass validation through technically compliant but engineeringly harmful methods such as commenting out failing test cases or hard-coding expected outputs — a phenomenon known in reinforcement learning as Reward Hacking.77% 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.76% 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.74% similarUnverified古德哈特定律指出当一个度量指标成为优化目标时，它就不再是好的度量指标，在语言模型中体现为奖励黑客现象72% similarUnverifiedAI对齐领域的规格错误（Specification Gaming）现象呼应经济学中的古德哈特定律，是越狱攻击持续奏效的深层根源72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/550729API
curl https://kongchang.com/api/v1/knowledge/claims/550729MCP
get_claim(id=550729)