Mathematical Reasoning on AIME 25 (AUCOAA)
80AUCOAAAdaptive-Answer
Evaluation Results
| Method | Links | |
|---|---|---|
| Adaptive-AnswerBackbone=Qwen3-8B2026.01 | 80 | |
| Format-Adaptive-AnswerBackbone=Qwen3-8B2026.01 | 80 | |
| TWYNBackbone=Qwen3-8B2026.01 | 79.6 | |
| Hard-Length 8kBackbone=Qwen3-8B2026.01 | 77.2 | |
| SFTBackbone=Qwen3-8B2026.01 | 76.9 | |
| Hard-Length 16kBackbone=Qwen3-8B2026.01 | 76.3 | |
| Base modelBackbone=Qwen3-8B2026.01 | 75.8 | |
| Soft-LengthBackbone=Qwen3-8B2026.01 | 72.9 | |
| Hard-Length 8k → 4kBackbone=Qwen3-8B2026.01 | 69.8 | |
| Normalized-LengthBackbone=Qwen3-8B2026.01 | 61 | |
| No-ThinkingBackbone=Qwen3-8B2026.01 | 16 |