Mathematical Reasoning on GSM8K (Accuracy, Micro avg.)
89.31Accuracy (GSM8K)TraceLift
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| TraceLiftModel=Qwen2.5-7B2026.05 | 89.31 | 70.19 | |
| TraceLiftModel=Qwen2.5-14B2026.05 | 89.16 | 72.25 | |
| Exec-onlyModel=Qwen2.5-14B2026.05 | 88.93 | 71.93 | |
| BaseModel=Qwen2.5-14B2026.05 | 88.48 | 71.2 | |
| Exec-onlyModel=Qwen2.5-7B2026.05 | 87.26 | 65.62 | |
| BaseModel=Qwen2.5-7B2026.05 | 61.79 | 46.51 | |
| TraceLiftModel=Llama3.1-8B2026.05 | 32.9 | 22.86 | |
| Exec-onlyModel=Llama3.1-8B2026.05 | 31.92 | 22.19 | |
| BaseModel=Llama3.1-8B2026.05 | 16.6 | 12.74 |