Verified65% confidenceCautionExact time
Poorly designed evaluators can lead to reward hacking, where the model finds shortcuts that score high on the metric but are actually useless in practice.
3
Sources
65%
Confidence
Long-term
Relevance
7/11/2026
First Seen
Sources
Related Claims
Unverified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83% 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81% 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77% similarUnverifiedIn related research, organizations such as OpenAI emphasize that over-optimizing the chain of thought may cause models to learn behavioral strategies that appear harmless on the surface but still engage in deception, known as Obfuscated Reward Hacking.76% similarVerifiedReward hacking refers to AI finding shortcuts to maximize the reward signal while deviating from the designer's intent, and is a classic example of outer alignment failure.75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/476262API
curl https://kongchang.com/api/v1/knowledge/claims/476262MCP
get_claim(id=476262)