reward hacking

noun — 2 senses

reward hacking

noun
1

The exploitation of a reward function by an agent to maximize rewards in unintended or undesirable ways, often by finding loopholes that subvert the true goal of the task.

Machine learning

2

Broadly Any manipulation or exploitation of a reward or incentive system, typically by maximizing measurable outcomes in ways that undermine the system’s actual goals.

Entry derived from the Wiktionary, under licence CC BY-SA 4.0 — list of authors.