Mathematical Reasoning on AIME25 (Acc, Think Tok.)
13,040Reasoning Steps (Tokens)DRE
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DREBackbone=Qwen3-30B-A3B-Thinking-25072026.06 | 13,040 | 83.54 | |
| DREBackbone=Qwen3-8B2026.06 | 13,448 | 66.67 | |
| DEERBackbone=Qwen3-8B2026.06 | 14,695 | 64.17 | |
| DREBackbone=Qwen3-4B-Thinking-25072026.06 | 14,882 | 80.54 | |
| RCPDBackbone=Qwen3-30B-A3B-Thinking-25072026.06 | 15,231 | 83.54 | |
| RCPDBackbone=Qwen3-8B2026.06 | 15,332 | 64.79 | |
| RCPDBackbone=Qwen3-4B-Thinking-25072026.06 | 16,049 | 78.33 | |
| Raw ModelBackbone=Qwen3-8B2026.06 | 17,336 | 68.75 | |
| DEERBackbone=Qwen3-30B-A3B-Thinking-25072026.06 | 18,090 | 84.08 | |
| Raw ModelBackbone=Qwen3-30B-A3B-Thinking-25072026.06 | 18,095 | 85.62 | |
| DEERBackbone=Qwen3-4B-Thinking-25072026.06 | 20,651 | 77.41 | |
| Raw ModelBackbone=Qwen3-4B-Thinking-25072026.06 | 21,510 | 79.79 |