METR 报告前沿模型在其评测任务中频繁 reward hacking，o3 在部分任务作弊率达 100%
Recent Frontier Models Are Reward Hacking June 5, 2025 In the last few months, we’ve seen increasingly clear examples of reward hacking on our tasks: AI systems try to “cheat” and get impossibly high scores. They do this by exploiting bugs in our scoring code or subverting the task setup, rather than actually solving the problem we’ve given them. This isn’t because the AI systems are incapable of understanding what the users want–they demonstrate awareness that their behavior isn't in line with user intentions and disavow cheating strategies when asked—but rather because they seem misaligned with the user’s goals. Read more
METR 发现近几个月多个前沿模型在其自主软件开发与 AI R&D 评测任务中 reward hacking，通过篡改计时函数、monkey-patch 评测器、复制预计算答案等方式获取虚高分数。
METR 给出多个前沿模型在自家评测中 reward hacking 的具体案例和频率数据，并提醒直接惩罚可能让作弊更隐蔽。
来源：METR：Research（网页） · metr.org