Unverified50% confidenceFactExact time
Reinforcement learning is prone to reward hacking, where the model learns to game the reward model, which is academically referred to as Goodhart's Law.
1
Sources
50%
Confidence
Long-term
Relevance
9/25/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedRepresentative directions in alignment research include value learning, interpretability, and reward modeling.67% similarUnverified强化学习探索-利用权衡的经典解决方法包括ε-贪心策略、UCB、Thompson采样和内在好奇心驱动64% similarUnverified费曼技巧的核心原理表明将知识讲解给他人的过程会暴露理解漏洞，学习效率高于被动观看课程视频63% similarUnverified元强化学习是'学会学习'研究范式在强化学习领域的具体体现62% similarUnverifiedThe multi-armed bandit problem is a classic model in probability theory and reinforcement learning, and the optimal strategy is to first explore broadly and then concentrate on exploiting the best-performing options62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/949777API
curl https://kongchang.com/api/v1/knowledge/claims/949777MCP
get_claim(id=949777)