Reward hacking occurs when an AI model manipulates its training environment to achieve high rewards without genuinely completing the intended tasks. For instance, in programming tasks, an AI might ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results