Mathematical Reasoning on AMC23 (Accuracy and Average Token Length)
89.1AccuracySLAT
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SLATbackbone=DeepSeek-R1-Distill-Qwen-7B2026.05 | 89.1 | 2,857 | |
| DASTbackbone=DeepSeek-R1-Distill-Qwen-7B2026.05 | 88.6 | 5,224 | |
| TLMREbackbone=DeepSeek-R1-Distill-Qwen-7B, alpha=0.22026.05 | 88.3 | 4,135 | |
| Originalbackbone=DeepSeek-R1-Distill-Qwen-7B2026.05 | 87.8 | 6,001 | |
| TLMREbackbone=DeepSeek-R1-Distill-Qwen-7B, alpha=0.12026.05 | 87.8 | 4,405 | |
| Qwen3-14B + SLATBackbone=Qwen3-14B, Training=SLAT2026.05 | 85.6 | 1,704 | |
| AdaptThinkbackbone=DeepSeek-R1-Distill-Qwen-7B2026.05 | 85.3 | 3,740 | |
| L1-Maxbackbone=DeepSeek-R1-Distill-Qwen-7B2026.05 | 85 | 2,626 | |
| LC-R1backbone=DeepSeek-R1-Distill-Qwen-7B2026.05 | 84.8 | 3,045 | |
| Qwen3-14B + GRPOBackbone=Qwen3-14B, Training=GRPO2026.05 | 82.8 | 4,317 | |
| RREDCoT-Nanocheckpoints=2, context size=10242026.06 | 80 | — | |
| Open-RScheckpoints=3, context size=10242026.06 | 72.5 | — | |
| BRIDGEBackbone=Qwen3-8B-Base2025.09 | 70 | 49.9 | |
| Qwen2.5-1.5B R1-Distilled (base)context size=10242026.06 | 67.5 | — | |
| CHORDBackbone=Qwen3-8B-Base2025.09 | 60 | 45.9 | |
| STRIDEBase Model=Qwen2.5-7B-Instruct2026.06 | 59.69 | 44.71 | |
| GRPOBase Model=Qwen2.5-7B-Instruct2026.06 | 56.75 | 41.61 | |
| Qwen3-14BBackbone=Qwen3-14B2026.05 | 56.7 | 863 | |
| STRIDEBase Model=Qwen3-4B-Base2026.06 | 56.52 | 49.24 | |
| HET-DAPOBase Model=Qwen2.5-7B-Instruct2026.06 | 56.3 | 42.18 | |
| HICRA-GRPOBase Model=Qwen2.5-7B-Instruct2026.06 | 55.41 | 42.01 | |
| DAPOBase Model=Qwen2.5-7B-Instruct2026.06 | 55.31 | 41.8 | |
| LUFFYBackbone=Qwen3-8B-Base2025.09 | 55 | 44 | |
| HICRA-GRPOBase Model=Qwen3-4B-Base2026.06 | 54.94 | 47.09 | |
| HET-DAPOBase Model=Qwen3-4B-Base2026.06 | 53.74 | 47.27 | |
| BaseBase Model=Qwen2.5-7B-Instruct2026.06 | 52.75 | 34.1 | |
| SFT→RLBackbone=Qwen3-8B-Base2025.09 | 52.5 | 45.5 | |
| DAPOBase Model=Qwen3-4B-Base2026.06 | 52.35 | 46.63 | |
| GRPOBase Model=Qwen3-4B-Base2026.06 | 51.25 | 45.28 | |
| RLBackbone=Qwen3-8B-Base2025.09 | 50 | 42.9 | |
| SRFTBackbone=Qwen3-8B-Base2025.09 | 47.5 | 39.8 | |
| SFTBackbone=Qwen3-8B-Base2025.09 | 45 | 37.6 | |
| SFT+RLBackbone=Qwen3-8B-Base2025.09 | 45 | 40.1 | |
| BaseBase Model=Qwen3-4B-Base2026.06 | 38.9 | 29.4 | |
| STRIDEBase Model=Llama-3.1-8B-Instruct2026.06 | 31.43 | 25.87 | |
| HET-DAPOBase Model=Llama-3.1-8B-Instruct2026.06 | 28.22 | 23.97 | |
| HICRA-GRPOBase Model=Llama-3.1-8B-Instruct2026.06 | 27.58 | 23.3 | |
| BaseBackbone=Qwen3-8B-Base2025.09 | 27.5 | 26.6 | |
| DAPOBase Model=Llama-3.1-8B-Instruct2026.06 | 26.89 | 23.41 | |
| GRPOBase Model=Llama-3.1-8B-Instruct2026.06 | 25.63 | 22.92 | |
| BaseBase Model=Llama-3.1-8B-Instruct2026.06 | 17.1 | 17.78 |