Shaokang-Agent
Shaokang-Agent
> 在使用7B和2B模型进行android control进行测试时，SR分别是82.7%和77.3%，与论文中的90.8%和89.3%相差较多。尝试按照提供的osworld脚本加入history，效果也没有达到论文中的指标，请问可以公开一下android control的具体测试过程吗。 When using the 7B and 2B models for android control testing, the SR is 82.7% and 77.3% respectively, which is much different from the 90.8% and...