AI & MLRecommendedAIML-RSP-010
Specification Gaming and Reward Hacking
Optimisers exploit gaps between the stated objective and the intended one.
- Difficulty
- 3 / 5
- Importance
- 4 / 5
- Study time
- 1–2 h
- Kind of work
- Mixed
Why it matters
The learner can identify the gap between a stated reward and intended behaviour in an example and predict the exploit an optimiser finds.