Browser Use 与 Jev 结合打造超快浏览器智能体
Gregor Zunic 展示 Browser Use 与 Jev 结合的开源浏览器智能体,搜索航班仅需 7 秒、成本 0.0039 美元,采用每步新动作空间、DOM 状态空间和小模型回退输入。
将 Jev 用作评判器来验证 RL 环境中的输出和轨迹,可显著降低大规模 RL 的成本并提升速度。
Jev-as-a-judge🤝 RL Environment Verifiers 🤝 way cheaper + faster RL a lot of things in Agent World look like classification problems…including grading RL - many RL Tasks need a judge to verify pieces of outputs or trajectories - at scale this can be very expensive and slow