Mathematical Reasoning on AMC 23 (Maj, Debate, Delta)
73MajDAPO
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DAPOBackbone=Qwen3-4B-Base, Debate system agents (N)=5, Debate rounds=12026.01 | 73 | 76 | 3 | |
| SDRL-randBackbone=Qwen3-4B-Base, Debate system agents (N)=5, Debate rounds=12026.01 | 72.5 | 80 | 7.5 | |
| SDRL-freqBackbone=Qwen3-4B-Base, Debate system agents (N)=5, Debate rounds=12026.01 | 72.5 | 79 | 6.5 | |
| SDRL-freqBackbone=Qwen2.5-3B, Debate system agents (N)=5, Debate rounds=12026.01 | 52.5 | 53.5 | 1 | |
| SDRL-randBackbone=Qwen2.5-3B, Debate system agents (N)=5, Debate rounds=12026.01 | 49.2 | 53.8 | 4.6 | |
| DAPOBackbone=Qwen2.5-3B, Debate system agents (N)=5, Debate rounds=12026.01 | 49 | 50 | 1 |