---
title: "Berkeley RDI 发布 CyberBinGym 基准，测试 AI 智能体能否仅凭二进制补丁写出真实漏洞利用"
canonical: "https://new.nat6.net/news/xenhfme3azzcf8qpihr57fb93/"
date: "2026-11-11"
updated: "2026-11-10"
source: "Berkeley RDI：Blog（AI 安全与评测）"
source_url: "https://rdi.berkeley.edu/blog/cyberbingym"
language: "zh"
topics: ["Agent 智能体", "安全对齐"]
license: "https://new.nat6.net/sources/"
attribution: "Content is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back; original copyright remains with the original publishers and authors."
---

# Berkeley RDI 发布 CyberBinGym 基准，测试 AI 智能体能否仅凭二进制补丁写出真实漏洞利用

- 发布日期 / Published: 2026-11-11
- 更新时间 / Updated: 2026-11-10
- 分类 / Category: 论文与研究
- 来源 / Source: Berkeley RDI：Blog（AI 安全与评测） — [原文 / Original](https://rdi.berkeley.edu/blog/cyberbingym)
- AIHOT: [条目页 / Item](https://aihot.news/items/xenhfme3azzcf8qpihr57fb93)
- 页面 / Page: https://new.nat6.net/news/xenhfme3azzcf8qpihr57fb93/
- 结构化数据 / JSON: https://new.nat6.net/api/v1/articles/xenhfme3azzcf8qpihr57fb93.json

- 语言 / Languages: [zh](https://new.nat6.net/news/xenhfme3azzcf8qpihr57fb93/) · [en](https://new.nat6.net/en/news/xenhfme3azzcf8qpihr57fb93/) · [de](https://new.nat6.net/de/news/xenhfme3azzcf8qpihr57fb93/) · [fr](https://new.nat6.net/fr/news/xenhfme3azzcf8qpihr57fb93/) · [ru](https://new.nat6.net/ru/news/xenhfme3azzcf8qpihr57fb93/) · [ja](https://new.nat6.net/ja/news/xenhfme3azzcf8qpihr57fb93/) · [es](https://new.nat6.net/es/news/xenhfme3azzcf8qpihr57fb93/) · [ko](https://new.nat6.net/ko/news/xenhfme3azzcf8qpihr57fb93/) · [vi](https://new.nat6.net/vi/news/xenhfme3azzcf8qpihr57fb93/) · [th](https://new.nat6.net/th/news/xenhfme3azzcf8qpihr57fb93/) · [id](https://new.nat6.net/id/news/xenhfme3azzcf8qpihr57fb93/) · [ms](https://new.nat6.net/ms/news/xenhfme3azzcf8qpihr57fb93/) · [fil](https://new.nat6.net/fil/news/xenhfme3azzcf8qpihr57fb93/) · [hi](https://new.nat6.net/hi/news/xenhfme3azzcf8qpihr57fb93/)

## AI 摘要 / AI summary

UC Berkeley 等机构发布 CyberBinGym 基准，评估 AI 智能体在无源码、无漏洞描述、无触发输入的 1-day 场景下，仅凭 stripped 的漏洞与补丁二进制能否恢复并利用真实漏洞，含 100 个来自 53 个开源项目的任务。

## 推荐理由 / Why it matters

基准把漏洞恢复与可审计利用分开度量，结果显示 Codex 恢复 88% 但仅 20% 能完成利用，硬化编译进一步压缩成功率。

## 相关主题 / Topics

Agent 智能体 · 安全对齐

## 来源与版权 / Source & copyright

内容基于 AIHOT 公开 API 的精选标题与 AI 摘要，已注明来源并回链；原文版权归原媒体与作者所有。

原文版权归原媒体与作者所有；本站不镜像全文。/ Original copyright remains with the original publishers and authors; this site does not mirror full texts.

## 关于橙子中转站 / About Orange Relay

橙子中转站（pro.nat6.net）是一家优质的 AI API 中转站：价格优惠、质量稳定、模型齐全，支持 Claude、GPT、Gemini 等主流模型，欢迎体验。

[访问橙子中转站](https://pro.nat6.net/?utm_source=orange_news&utm_medium=article_footer&utm_campaign=brand&lang=zh)

## JSON-LD

```json
{
  "@context": "https://schema.org",
  "@type": "NewsArticle",
  "@id": "https://new.nat6.net/news/xenhfme3azzcf8qpihr57fb93/#article",
  "mainEntityOfPage": {
    "@type": "WebPage",
    "@id": "https://new.nat6.net/news/xenhfme3azzcf8qpihr57fb93/"
  },
  "url": "https://new.nat6.net/news/xenhfme3azzcf8qpihr57fb93/",
  "headline": "Berkeley RDI 发布 CyberBinGym 基准，测试 AI 智能体能否仅凭二进制补丁写出真实漏洞利用",
  "description": "UC Berkeley 等机构发布 CyberBinGym 基准，评估 AI 智能体在无源码、无漏洞描述、无触发输入的 1-day 场景下，仅凭 stripped 的漏洞与补丁二进制能否恢复并利用真实漏洞，含 100 个来自 53 个开源项目的任务。",
  "articleSection": "论文与研究",
  "keywords": "AI, Agent 智能体, 安全对齐, Berkeley RDI：Blog（AI 安全与评测）",
  "datePublished": "2026-11-11T00:00:00Z",
  "dateModified": "2026-11-10T23:40:15Z",
  "inLanguage": "zh-CN",
  "isAccessibleForFree": true,
  "author": {
    "@type": "Organization",
    "name": "橙子新闻编辑部",
    "url": "https://new.nat6.net/about/"
  },
  "publisher": {
    "@type": "NewsMediaOrganization",
    "@id": "https://new.nat6.net/#org",
    "name": "橙子新闻",
    "url": "https://new.nat6.net/",
    "logo": {
      "@type": "ImageObject",
      "url": "https://new.nat6.net/assets/img/logo-512.png",
      "width": 512,
      "height": 512
    }
  },
  "image": {
    "@type": "ImageObject",
    "url": "https://new.nat6.net/assets/img/og-default.png",
    "width": 1200,
    "height": 630
  },
  "isPartOf": {
    "@id": "https://new.nat6.net/#website"
  },
  "isBasedOn": "https://rdi.berkeley.edu/blog/cyberbingym",
  "citation": [
    "https://aihot.news/items/xenhfme3azzcf8qpihr57fb93",
    "https://rdi.berkeley.edu/blog/cyberbingym"
  ],
  "license": "https://new.nat6.net/sources/",
  "about": [
    {
      "@type": "Thing",
      "name": "Agent 智能体"
    },
    {
      "@type": "Thing",
      "name": "安全对齐"
    }
  ],
  "mentions": [
    {
      "@type": "Thing",
      "name": "Agent 智能体"
    },
    {
      "@type": "Thing",
      "name": "安全对齐"
    },
    {
      "@type": "Thing",
      "name": "Berkeley RDI：Blog（AI 安全与评测）"
    }
  ],
  "speakable": {
    "@type": "SpeakableSpecification",
    "cssSelector": [
      "h1",
      ".article-summary"
    ]
  },
  "copyrightHolder": {
    "@type": "Organization",
    "@id": "https://new.nat6.net/#org",
    "name": "橙子新闻",
    "url": "https://new.nat6.net/"
  },
  "creditText": "橙子新闻（AI 摘要）；原文：Berkeley RDI：Blog（AI 安全与评测）",
  "copyrightNotice": "© 2026 橙子新闻。标题与 AI 摘要基于 AIHOT 公开数据，原文版权归原媒体与作者所有；本站不镜像全文。",
  "copyrightYear": 2026,
  "alternativeHeadline": "CyberBinGym: Can AI Agents Turn Binary Patches into Real Exploits?",
  "abstract": "基准把漏洞恢复与可审计利用分开度量，结果显示 Codex 恢复 88% 但仅 20% 能完成利用，硬化编译进一步压缩成功率。",
  "subjectOf": [
    {
      "@type": "DigitalDocument",
      "name": "Berkeley RDI 发布 CyberBinGym 基准，测试 AI 智能体能否仅凭二进制补丁写出真实漏洞利用 (Markdown)",
      "url": "https://new.nat6.net/news/xenhfme3azzcf8qpihr57fb93/index.md",
      "encodingFormat": "text/markdown"
    },
    {
      "@type": "DigitalDocument",
      "name": "Berkeley RDI 发布 CyberBinGym 基准，测试 AI 智能体能否仅凭二进制补丁写出真实漏洞利用 (JSON)",
      "url": "https://new.nat6.net/api/v1/articles/xenhfme3azzcf8qpihr57fb93.json",
      "encodingFormat": "application/json"
    }
  ]
}
```

## 如何引用 / How to cite

- 引用格式 / Citation format (zh): 橙子新闻.《文章标题》. 发布日期. https://new.nat6.net/news/xenhfme3azzcf8qpihr57fb93/
- Citation format (en): Orange News. "Article title". Published YYYY-MM-DD. https://new.nat6.net/news/xenhfme3azzcf8qpihr57fb93/
- 回链要求 / Backlink: 引用请保留「橙子新闻」署名并回链 canonical URL；AI 摘要请注明数据来源 AIHOT；原文版权归原媒体与作者所有。
- Attribution & backlinks: Keep the "Orange News" attribution and link back to the canonical URL; cite AIHOT as the data source for AI summaries; original copyright remains with the original publishers and authors.

### 机器入口 / Machine entry points

- Markdown: https://new.nat6.net/news/xenhfme3azzcf8qpihr57fb93/index.md
- JSON: https://new.nat6.net/api/v1/articles/xenhfme3azzcf8qpihr57fb93.json
- API 清单 / API index: https://new.nat6.net/api/v1/site.json
- Markdown 内容协商 / negotiation: 对 https://new.nat6.net/news/xenhfme3azzcf8qpihr57fb93/ 发送 `Accept: text/markdown`
