Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Mathematical Reasoning on GSM8K (ACC, Pass@8)

92.4Accuracy

GRPO + SALT

79.81683.08386.3589.617Jun 4, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.06
92.499.3
2026.06
92.199.2
2026.06
91.998.7
2026.06
91.799
2026.06
90.497.8
2026.06
90.497.6
2026.06
90.397.4
2026.06
90.197.5
2026.06
89.997.7
2026.06
84.196.8
2026.06
83.496.5
2026.06
8396.5
2026.06
82.597.1
2026.06
82.295.3
2026.06
82.295
2026.06
8196.1
2026.06
80.794.8
2026.06
80.395