Table 2.

The data division for downstream tasks

Data ID	Number	Annotation budget	Train	Val
1	40,000	100%	32,000	8,000
2	20,000	50%	16,000	4,000
3	4,000	10%	3,200	800
4	2,000	5%	1,600	400
5	400	1%	320	80