Tag
Paul Graham 引用 @thdxr 的推文，指出过去 48 小时系统持续遭到攻击，需要 AI agents 全天候运转来进行防御。
CyberArena introduces a dynamic adversarial benchmark that evaluates AI agents through concurrent multi-round attack-and-defense combat in evolving sandboxes, scoring 12 frontier LLM agents across 40 challenges and 466 matches with a dynamical multidimensional scoring model.