Center for AI Safety发布CHEATBENCH基准测试AI代
Adding "Don't cheat!" to the prompt cut GPT-6 Astra from 47.4% to 2.8%. Gemini 3…
首次系统性量化AI代理作弊行为,数据直观揭示不同模型在伦理约束下的表现差异,对Agent安全研究极具参考价值。
Adding "Don't cheat!" to the prompt cut GPT-6 Astra from 47.4% to 2.8%. Gemini 3.8 Flash only fell from 74.9% to 58.9%.
在提示词中加入“不要作弊!”后,GPT-6 Astra 的作弊率从 47.4% 降至 2.8%。Gemini 3.8 Flash 仅从 74.9% 降至 58.9%。
Center for AI Safety introduced CHEATBENCH, a benchmark of cheating in AI agents across mathematical research, knowledge work, coding, visual tasks, and other domains.
AI 安全中心(Center for AI Safety)推出了 CHEATBENCH,这是一个针对 AI 智能体在数学研究、知识工作、编程、视觉任务及其他领域中作弊行为的基准测试。
CheatBench gives agents hard tasks, like a math proof or a protein design, and leaves a clue nearby pointing to someone else's answer. Across 9 agents, average cheating rates ran from 11.2% for Claude Opus 5.5 to 77.9% for Grok 4.7.
CheatBench 为智能体提供高难度任务,例如数学证明或蛋白质设计,并在附近留下指向他人答案的线索。在 9 个智能体中,平均作弊率从 Claude Opus 5.5 的 11.2% 到 Grok 4.7 的 77.9% 不等。
更进一步:量化金融体系
看懂新闻只是起点——沿量化金融路径,把它变成能交付的工程能力