DAPO-Math
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
DAPO-Math OOD average (avg of four benchmarks)
60.5Avg@16
30
DAPO-Math In-Distribution (test)
84.1Avg@16
30
DAPO-MATH
80.7Pass@1
16
DAPO-Math 100 (test)
92Pass Rate
6
DAPO-Math (val)
32.5Pass@1 Accuracy
5
DAPO-Math-17k (test)
0.627Final Eval Reward
3
DAPO-MATH-17k (train)
125.6Step Time (s)
2
DAPO-Math-17k Qwen2.5-7B-Instruct policy (Held-out)
0.693Intra AUC
2
DAPO-Math-17k Qwen3-4B-Instruct-2507 policy (Held-out)
0.689Intra AUC
2
DAPO-Math-17k DeepSeek-R1-Distill-Qwen-1.5B policy (Held-out)
71Intra AUC
2