InternVL icon indicating copy to clipboard operation
InternVL copied to clipboard

Consistency training和Router training用了多少数据量?

Open whu125 opened this issue 2 months ago • 0 comments

论文中提到Consistency training复用了SFT的数据,请问是完全一致的数据量吗?还是挑选了部分子集呢?

同理,Router training具体使用了多少呢?

whu125 avatar Oct 14 '25 06:10 whu125