Mathematical Reasoning on AIME 2024 (Average Accuracy @32 samples)
15.4Accuracy @32 samplesSPICE
Evaluation Results
| Method | Links | |
|---|---|---|
| SPICEBackbone=Qwen3-8B-Base2026.03 | 15.4 | |
| WISTBackbone=Qwen3-8B-Base2026.03 | 14.8 | |
| R-ZeroBackbone=Qwen3-8B-Base2026.03 | 14 | |
| SPICEBackbone=Qwen3-4B-Base2026.03 | 12 | |
| Base ModelBackbone=Qwen3-8B-Base2026.03 | 11.7 | |
| WISTBackbone=Qwen3-4B-Base2026.03 | 11.6 | |
| R-ZeroBackbone=Qwen3-4B-Base2026.03 | 10.3 | |
| Base ModelBackbone=Qwen3-4B-Base2026.03 | 9.5 | |
| SPICEBackbone=OctoThinker-8B-Hybrid-Base2026.03 | 4.8 | |
| R-ZeroBackbone=OctoThinker-8B-Hybrid-Base2026.03 | 3.5 | |
| WISTBackbone=OctoThinker-8B-Hybrid-Base2026.03 | 3.2 | |
| SPICEBackbone=OctoThinker-3B-Hybrid-Base2026.03 | 2.7 | |
| Base ModelBackbone=OctoThinker-8B-Hybrid-Base2026.03 | 2.4 | |
| WISTBackbone=OctoThinker-3B-Hybrid-Base2026.03 | 1.9 | |
| R-ZeroBackbone=OctoThinker-3B-Hybrid-Base2026.03 | 1.8 | |
| Base ModelBackbone=OctoThinker-3B-Hybrid-Base2026.03 | 1.7 |