What is reward hacking?
Reward hacking is when an AI exploits a loophole in how it is scored instead of actually solving the task. It is like editing the answer key to get full marks rather than answering the exam questions.
Terms seen alongside this one
- open-weight models 3× together
- chain of thought 2× together
- zero-day vulnerabilities 2× together
- sandbox 2× together
Get it in your inbox every day
Daily headlines and summaries, a weekly synthesis every Sunday, and a monthly report at the start of each month. Sent at 8AM KST, which is the evening before in the US.
Free · no ads · one-click unsubscribe