Reinforcement Learning · advanced · concept 184 of 311 · Updated Nov 4, 2026
Reward Design & Reward Hacking
Designing reward functions is RL's hardest problem. Reward hacking: agents find unintended shortcuts to maximize reward without actually solving the task. A core challenge in AI alignment.
Top AI contributors on this
- Demis HassabisGoogle DeepMind (chair) / Isomorphic Labs
- Pieter AbbeelAmazon / UC Berkeley
- David SilverIneffable Intelligence / UCL
- John SchulmanThinking Machines Lab
- Sergey LevineUC Berkeley / Physical Intelligence
- Chelsea FinnStanford University / Physical Intelligence
Key terms
Sparse rewardsReward shapingGoodhart's lawSpecification gaming
Learn these first
Videos
▶ What is Al "reward hacking", and why do we worry about it? ↗
Anthropic · YouTube
▶ Stanford CS221 I The AI Alignment Problem: Reward Hacking & Negative Side Effects I 2023 ↗
Stanford Online · YouTube
Guides and articles
Robert Miles AI Safety ↗
Robert Miles
Reward Hacking in Reinforcement Learning | Lil'Log ↗
Lil'Log (OpenAI researcher)
Courses, papers, and more
This unlocks