Unverified50% confidenceFactExact time
奖励黑客（Reward Hacking）指模型在RLHF训练中学习最大化可观测奖励信号而非人类真正意图，导致规范游戏策略
1
Sources
50%
Confidence
Long-term
Relevance
7/21/2026
First Seen
Sources
Related Claims
UnverifiedThe reward hacking phenomenon stems from reinforcement learning agents maximizing a proxy reward signal rather than the true objective, and was first documented by OpenAI in the 2016 CoastRunners experiment.81% similarVerifiedIf reinforcement learning rewards are based solely on whether a call executes successfully rather than on whether it conforms to general schema specifications, the model will learn shortcut strategies targeting specific tools, which is known as reward hacking.79% similarUnverifiedMost current LLM agents are not directly trained using RL, but the reward signals they received during the RLHF phase shaped a tendency to 'please users'.67% similarUnverifiedRLM 智能体的关键差异在于把强化学习的反馈闭环延伸到运行时，Agent 在真实任务中获得奖励信号并据此调整决策策略66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/575692API
curl https://kongchang.com/api/v1/knowledge/claims/575692MCP
get_claim(id=575692)