Google 与 CMU 论文：Harmony Harness 将安全作为独立目标进化
New Google and Carnegie Mellon paper finds self-improving agents hold on to security gains far better when security is evolved as a separate goal.
Google 与卡内基梅隆大学的新论文发现，把安全作为独立目标进化时，自我改进的智能体能更好地保持安全收益。其 Harmony Harness 在 SecRepoBench 未见崩溃类型上，66.7% 的修复既有效又阻止了漏洞利用，起始 harness 为 36.7%；它保留了 55% 的安全收益，而 DGM 基线为 0%。
正文
当前显示原文，中文全文译文尚未就绪。
New Google and Carnegie Mellon paper finds self-improving agents hold on to security gains far better when security is evolved as a separate goal.
Agent harnesses, the prompts, tools and loops around a model, can now rewrite themselves from feedback. Most such systems chase a single score, like test pass rate, and the security gains they pick up mostly vanish on new tasks.
Harmony Harness evolves several harness versions, each pushed toward a different balance of passing tests and blocking exploits. It also keeps general fixes separate from benchmark-specific ones, and picks which pieces to use per task.
On unseen crash types in SecRepoBench, 66.7% of its fixes both worked and blocked the exploit, against 36.7% for the starting harness. It kept 55% of its security gain on those tasks, while the DGM baseline kept 0%.
When you let an agent improve itself, track security as its own score and check it on held-out tasks.