Mathematical Reasoning on MATH 500 (Accuracy and Thinking Token Usage)
88.5AccuracyDART
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DARTModel Scale=32B, Thinking Strategy=DART2026.06 | 88.5 | 69 | |
| DARTModel Scale=8B, Thinking Strategy=DART2026.06 | 88.2 | 67 | |
| ATModel Scale=14B, Thinking Strategy=AT2026.06 | 87.6 | — | |
| DARTModel Scale=14B, Thinking Strategy=DART2026.06 | 87.6 | 37 | |
| ATModel Scale=32B, Thinking Strategy=AT2026.06 | 86.2 | — | |
| ATModel Scale=8B, Thinking Strategy=AT2026.06 | 85.6 | — | |
| NTModel Scale=32B, Thinking Strategy=NT2026.06 | 82.2 | — | |
| NTModel Scale=14B, Thinking Strategy=NT2026.06 | 81.2 | — | |
| NTModel Scale=8B, Thinking Strategy=NT2026.06 | 76.6 | — | |
| ATModel Scale=4B, Thinking Strategy=AT2026.06 | 70.5 | — | |
| NTModel Scale=4B, Thinking Strategy=NT2026.06 | 66.5 | — | |
| ATModel Scale=1.7B, Thinking Strategy=AT2026.06 | 65.5 | — | |
| DARTModel Scale=4B, Thinking Strategy=DART2026.06 | 65.5 | 30 | |
| DARTModel Scale=1.7B, Thinking Strategy=DART2026.06 | 58 | 29 | |
| ATModel Scale=0.6B, Thinking Strategy=AT2026.06 | 56 | — | |
| DARTModel Scale=0.6B, Thinking Strategy=DART2026.06 | 54 | 23 | |
| NTModel Scale=1.7B, Thinking Strategy=NT2026.06 | 52.5 | — | |
| NTModel Scale=0.6B, Thinking Strategy=NT2026.06 | 38.5 | — |