Data Pool Size
1024M
256M
512M
1024M
1B
2B
3B
4B
5B
40%
45%
50%
55%
60%
DatologyAI Classification-Optimized (Pool Size: 1024M)
Sophisticated Baseline (Pool Size: 1024M)
Raw Baseline (Pool Size: 1024M)
Classification-Optimized Curation
Substantially Improves Classification Performance
Total Training Samples
Final Training Accuracy