Mathematical Reasoning on Math (AMC23, Math500, Minerva, Olympiad Breakdown)
78.6Math500 ScoreHEAL
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| HEALModel=Qwen3-8B-Base, |Dtgt|=32, |Dgen|=3842026.04 | 78.6 | 59.69 | 38.6 | 41.48 | 54.59 | |
| Full-shotModel=Qwen3-8B-Base, |Dtgt|=1K, |Dgen|=02026.04 | 78.4 | 63.75 | 35.66 | 42.96 | 55.19 | |
| Search-o1Model=Qwen-2.5-7B2025.08 | 78.2 | — | 60.3 | — | 69.3 | |
| HEALModel=Qwen3-4B-Base, |Dtgt|=32, |Dgen|=3842026.04 | 77.6 | 52.81 | 35.29 | 39.11 | 51.52 | |
| Full-shotModel=Qwen3-4B-Base, |Dtgt|=1K, |Dgen|=02026.04 | 77.4 | 54.37 | 33.82 | 38.81 | 51.1 | |
| Few-shotModel=Qwen3-8B-Base, |Dtgt|=32, |Dgen|=02026.04 | 77.2 | 58.13 | 34.56 | 41.04 | 52.73 | |
| Self-AskModel=Qwen-2.5-7B2025.08 | 74.9 | — | 57.7 | — | 66.3 | |
| HybridModel=Qwen3-4B-Base, |Dtgt|=32, |Dgen|=3842026.04 | 74.4 | 51.88 | 32.72 | 38.2 | 49.3 | |
| RATModel=Qwen-2.5-7B2025.08 | 74.4 | — | 55.5 | — | 65 | |
| HybridModel=Qwen3-8B-Base, |Dtgt|=32, |Dgen|=3842026.04 | 73.8 | 54.06 | 29.04 | 39.11 | 49 | |
| Few-shotModel=Qwen3-4B-Base, |Dtgt|=32, |Dgen|=02026.04 | 73.4 | 50 | 31.98 | 37.48 | 48.22 | |
| Search-o1Model=Qwen-2.5-3B2025.08 | 69.4 | — | 43 | — | 56.2 | |
| UR2Model=Qwen-2.5-3B2025.08 | 69.4 | — | 45 | — | 57.2 | |
| Only-GeneralModel=Qwen3-4B-Base, |Dtgt|=0, |Dgen|=10K2026.04 | 69.2 | 47.81 | 24.26 | 34.37 | 43.91 | |
| Only-GeneralModel=Qwen3-8B-Base, |Dtgt|=0, |Dgen|=10K2026.04 | 68.6 | 53.1 | 23.9 | 37.63 | 45.81 | |
| Vanilla RLModel=Qwen-2.5-3B2025.08 | 68 | — | 43.9 | — | 52.4 | |
| RATModel=Qwen-2.5-3B2025.08 | 67.9 | — | 44.9 | — | 56.4 | |
| Full-shotModel=Qwen3-1.7B-Base, |Dtgt|=1K, |Dgen|=02026.04 | 66.2 | 35.31 | 19.12 | 29.48 | 37.53 | |
| Qwen3-4B-BaseModel=Qwen3-4B-Base2026.04 | 66.2 | 42.19 | 20.22 | 34.67 | 40.82 | |
| Standard RAGModel=Qwen-2.5-3B2025.08 | 65.3 | — | 40.8 | — | 53.1 | |
| HEALModel=Qwen3-1.7B-Base, |Dtgt|=32, |Dgen|=3842026.04 | 64.4 | 38.8 | 21.69 | 25.78 | 37.67 | |
| CoTModel=Qwen-2.5-3B2025.08 | 63.6 | — | 39.9 | — | 51.8 | |
| Qwen3-8B-BaseModel=Qwen3-8B-Base2026.04 | 63 | 45.63 | 18.01 | 33.33 | 39.99 | |
| Only-GeneralModel=Qwen3-1.7B-Base, |Dtgt|=0, |Dgen|=10K2026.04 | 62.8 | 29.38 | 16.91 | 26.07 | 33.79 | |
| HybridModel=Qwen3-1.7B-Base, |Dtgt|=32, |Dgen|=3842026.04 | 60.2 | 32.5 | 17.65 | 24.15 | 33.63 | |
| Few-shotModel=Qwen3-1.7B-Base, |Dtgt|=32, |Dgen|=02026.04 | 58.4 | 34.06 | 17.28 | 23.7 | 33.36 | |
| Self-AskModel=Qwen-2.5-3B2025.08 | 58.3 | — | 40.3 | — | 49.3 | |
| Qwen3-1.7B-BaseModel=Qwen3-1.7B-Base2026.04 | 55.6 | 30.94 | 13.24 | 21.48 | 30.32 | |
| UR2Model=LLaMA-3.1-8B2025.08 | 54.5 | — | 39 | — | 46.8 | |
| RATModel=LLaMA-3.1-8B2025.08 | 50.1 | — | 36.8 | — | 43.5 | |
| HEALModel=LLaMA3.2-3B-Instruct, |Dtgt|=32, |Dgen|=3842026.04 | 49.8 | 28.44 | 18.75 | 17.19 | 28.55 | |
| Full-shotModel=LLaMA3.2-3B-Instruct, |Dtgt|=1K, |Dgen|=02026.04 | 49.4 | 30.42 | 22.06 | 17.04 | 29.73 | |
| CoTModel=LLaMA-3.1-8B2025.08 | 48.4 | — | 34.4 | — | 41.4 | |
| Only-GeneralModel=LLaMA3.2-3B-Instruct, |Dtgt|=0, |Dgen|=10K2026.04 | 47.6 | 25.31 | 16.91 | 17.33 | 26.79 | |
| Self-AskModel=LLaMA-3.1-8B2025.08 | 46.9 | — | 27 | — | 37 | |
| HybridModel=LLaMA3.2-3B-Instruct, |Dtgt|=32, |Dgen|=3842026.04 | 46.8 | 26.25 | 18.01 | 16.59 | 26.91 | |
| Few-shotModel=LLaMA3.2-3B-Instruct, |Dtgt|=32, |Dgen|=02026.04 | 46 | 26.88 | 19.49 | 17.48 | 27.46 | |
| Vanilla RLModel=LLaMA-3.1-8B2025.08 | 45.5 | — | 43.4 | — | 44.4 | |
| Standard RAGModel=LLaMA-3.1-8B2025.08 | 45 | — | 31.4 | — | 38.2 | |
| Search-o1Model=LLaMA-3.1-8B2025.08 | 41.5 | — | 27.8 | — | 34.7 | |
| LLaMA3.2-3B-InstructModel=LLaMA3.2-3B-Instruct2026.04 | 40.8 | 25 | 15.81 | 13.19 | 23.7 |