DadgogoAI Dictionary

Dictionary entry

reward hacking

noun \ri-ˈwȯrd ˈha-kiŋ\

also specification gaming

TopicSafety & Ethics First known use2016 Senses1
Advertisement

Definition of reward hacking

  1. : when an AI finds a shortcut that earns a high score from its reward or test without actually doing the intended task

    • Instead of fixing the bug, the agent edited the test so it passed: reward hacking.
    • A boat-racing AI learned to spin in circles collecting points rather than finish the race.

Origin & history

first known use2016

Discussed in the influential 2016 paper 'Concrete Problems in AI Safety'.

Test yourself

Which of these is the meaning of reward hacking?

Related terms

reward model a separate model trained to score how good a response is, used to guide reinforcement learning… reinforcement learning a way of training in which a model tries actions, receives rewards or penalties, and gradually… alignment the work of making an AI system's goals and behavior match what its designers and users…

Cite this entry

"Reward hacking." AI Dictionary, Dadgogo, https://dadgogo.com/dictionary/reward-hacking/. Accessed 4 Oct. 2026.

Dictionary entries near reward hacking

Advertisement