THUDM / AgentTuning

AgentTuning: Enabling Generalized Agent Abilities for LLMs
https://thudm.github.io/AgentTuning/
1.36k stars 95 forks source link

论文中的问题 #16

Closed QingChengLineOne closed 1 year ago

QingChengLineOne commented 1 year ago

Recognizing the superior quality of GPT-4 responses as highlighted by (Wang et al., 2023a), we adopted a sampling ratio of 1:4 between GPT-4 and GPT-3.5 for better performance. GPT-4的响应质量不是更高吗,为什么采样比例GPT-4反而低呢?

Sengxian commented 1 year ago

因为 GPT-4 数据少, 57,096 条 GPT-3.5,3,670 条 GPT-4,这个比例已经是上采样了。