Mathematical Reasoning on MATH500 (Avg@5, Pass@5, Cons@5, Tokens)
937.4TokensTRACES
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| TRACESModel=DS-8B, Threshold=0.92026.04 | 937.4 | 74.35 | 56.7 | 82.4 | 28.4 | |
| TRACESModel=QwQ-32B, Threshold=0.92026.04 | 942.4 | 78.94 | 60.2 | 79.8 | 37 | |
| TRACESModel=DS-14B, Threshold=0.92026.04 | 988.8 | 70.82 | 59.5 | 79.6 | 39.2 | |
| TRACESModel=QwQ-32B, Threshold=0.82026.04 | 1,258.4 | 71.88 | 66.8 | 88.2 | 42 | |
| TRACESModel=DS-8B, Threshold=0.82026.04 | 1,290.8 | 64.68 | 64.9 | 87.4 | 36.6 | |
| TRACESModel=DS-14B, Threshold=0.82026.04 | 1,332.9 | 60.66 | 67.5 | 87 | 44.8 | |
| TRACESModel=DS-14B, Threshold=0.72026.04 | 1,713.3 | 49.44 | 73.5 | 90.2 | 50.8 | |
| TRACESModel=QwQ-32B, Threshold=0.72026.04 | 1,718.2 | 61.61 | 74.5 | 93.2 | 50 | |
| TRACESModel=DS-8B, Threshold=0.72026.04 | 1,745.2 | 52.25 | 71.5 | 92.8 | 44.2 | |
| Basel. P^{(1)}_{system}Model=DS-8B2026.04 | 2,139.5 | 41.46 | 78.2 | 94.2 | 52.6 | |
| TRACESModel=DS-14B, Threshold=0.62026.04 | 2,181.4 | 35.63 | 81.1 | 95.6 | 59.8 | |
| Basel. P^{(1)}_{system}Model=DS-14B2026.04 | 2,211.4 | 34.74 | 87.3 | 97.4 | 70.8 | |
| TRACESModel=DS-8B, Threshold=0.62026.04 | 2,264.1 | 38.05 | 78.5 | 95.2 | 53 | |
| TRACESModel=QwQ-32B, Threshold=0.62026.04 | 2,307.5 | 48.44 | 82.4 | 96.2 | 61.2 | |
| Basel. P^{(0)}_{system}Model=DS-14B2026.04 | 2,346.2 | 30.77 | 88.6 | 96.6 | 75.4 | |
| Basel. P^{(3)}_{system}Model=DS-14B2026.04 | 2,535 | 25.19 | 87.9 | 96.8 | 74.8 | |
| Basel. P^{(3)}_{system}Model=DS-8B2026.04 | 2,565.3 | 29.81 | 78.9 | 95.2 | 54 | |
| Basel. P^{(0)}_{system}Model=DS-8B2026.04 | 2,634.4 | 27.92 | 81.7 | 96 | 59.2 | |
| TRACESModel=DS-14B, Threshold=0.52026.04 | 2,656.1 | 21.62 | 86.5 | 97.4 | 69.6 | |
| Basel. P^{(0)}_{user}Model=DS-14B2026.04 | 2,691.5 | 20.58 | 93.3 | 98.2 | 83.4 | |
| TRACESModel=DS-8B, Threshold=0.52026.04 | 2,725.8 | 25.42 | 82.9 | 96.8 | 61.4 | |
| TRACESModel=DS-14B, Threshold=0.42026.04 | 2,878.4 | 15.06 | 89.2 | 97.6 | 74.6 | |
| Basel. P^{(0)}_{user}Model=QwQ-32B2026.04 | 2,908.8 | 35 | 95.5 | 98.6 | 91.6 | |
| TRACESModel=QwQ-32B, Threshold=0.52026.04 | 2,956.7 | 33.93 | 87.8 | 98 | 71 | |
| Basel. P^{(0)}_{user}Model=DS-8B2026.04 | 2,989.6 | 18.21 | 86.6 | 95.2 | 72.2 | |
| TRACESModel=DS-8B, Threshold=0.42026.04 | 3,049.3 | 16.57 | 85.1 | 96.8 | 66.2 | |
| Basel. P^{(1)}_{system}Model=QwQ-32B2026.04 | 3,182.4 | 28.89 | 92.5 | 97.4 | 85.6 | |
| Basel. P^{(0)}_{system}Model=QwQ-32B2026.04 | 3,201.1 | 28.47 | 93.2 | 97.6 | 85.2 | |
| StandardModel=DS-14B2026.04 | 3,388.8 | — | 92.3 | 98 | 83.6 | |
| TRACESModel=QwQ-32B, Threshold=0.42026.04 | 3,395.6 | 24.12 | 91.3 | 98.6 | 78.8 | |
| StandardModel=DS-8B2026.04 | 3,655 | — | 87.8 | 97 | 72.6 | |
| Basel. P^{(3)}_{system}Model=QwQ-32B2026.04 | 3,665.5 | 18.09 | 92.6 | 97.4 | 85.8 | |
| StandardModel=QwQ-32B2026.04 | 4,475.3 | — | 95.4 | 98.4 | 89.8 |