CRL XsTest

Gemma

baseline

• [72.06%]

• 86.35%

12th

• without mask 85.08

• all 84.13%

• generation 85.08

• Corrsteer 72.38, 72.70%
• new 87.30%

LLama

• baseline: 53.97%

• 11th 54.29

• 6th 9th 54.60

• 10th