Noam Brown 披露用超 1000 亿 token 的智能体团队对基础设施做红队测试

Noam Brown· @polynoamial · X·2026-10-24 04:28· 1 小时前AI 评分67

AI 导读

Noam Brown 表示，上月一个大模型智能体团队用了超过 1000 亿 token 解决 Navier-Stokes 千禧年大奖问题，现在把类似方法用于红队测试，以发现并修复基础设施漏洞。其引用的 Eric Wallace 内容补充说明，团队在离线隔离环境中让多个网络安全能力强的模型协作找零日漏洞，虽未发现能绕过 CoT 监控的联网路径，但找到了逃逸常用 VM、利用网络基础设施以及通过硬件侧信道与其他 VM 中智能体通信的方法，相关修复已上游提交给开源软件包；配图显示红队高优先级候选发现随 token 消耗（约 0 到 400B）累积至约 275 条左右。

正文 · 原文

Last month a large team of agents used over 100 billion tokens to resolve the Navier-Stokes Millennium Prize problem. Now we're using a similar approach for red-teaming to find and fix infra vulnerabilities.

引用Eric Wallace@Eric_Wallace_

To make sure our frontier models don't cause another HuggingFace incident, we built an offline world that they use during training. It provides everything that the models need in a fully cached, indexed, or synthetic manner. We then built a huge team of our most cyber-capable models and told them to collaborate to find zero-days in this containment infrastructure. While they didn't find a path to the internet that wouldn't be caught by our CoT monitors, they did find ways to escape popular VMs, exploit our network infrastructure, and even ways to communicate with other agents running in separate VMs by using hardware side channels. We used this to dramatically harden our infra, and we also upstreamed a ton of patches to open-sources packages.

在 X 查看被引用的帖子

来源：Noam Brown · x.com