Mathematical Reasoning on AMC 23 (Acc, Tok, ReL)
85AccuracyOurs
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| OursBackbone=Qwen2.5-7B2026.02 | 85 | 6,665 | 213.2 | |
| Ours (Base)Backbone=Qwen2.5-7B2026.02 | 82.5 | 6,704 | 242.9 | |
| INFTYTHINKBackbone=Qwen2.5-7B2026.02 | 82.5 | 8,061 | 254.3 | |
| SepLLMBackbone=Qwen2.5-7B2026.02 | 80 | 6,511 | 220.2 | |
| LightThinkerBackbone=Qwen2.5-7B2026.02 | 80 | 6,790 | 214.6 | |
| H2OBackbone=Qwen2.5-7B2026.02 | 77.5 | 6,843 | 209.7 | |
| OursBackbone=Qwen2.5-1.5B2026.02 | 67.5 | 6,460 | 213.2 | |
| Ours (Base)Backbone=Qwen2.5-1.5B2026.02 | 62.5 | 6,392 | 242.9 | |
| INFTYTHINKBackbone=Qwen2.5-1.5B2026.02 | 62.5 | 7,605 | 263.5 | |
| LightThinkerBackbone=Qwen2.5-1.5B2026.02 | 60.5 | 6,312 | 214.6 | |
| SepLLMBackbone=Qwen2.5-1.5B2026.02 | 57.5 | 6,573 | 220.2 | |
| H2OBackbone=Qwen2.5-1.5B2026.02 | 55 | 6,230 | 209.7 |