Mathematical Reasoning on DART
46.8AccuracyQwen-7B-Instruct
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen-7B-InstructRouter=true2025.10 | 46.8 | 273.1 | |
| Qwen-7B-InstructRouter=false2025.10 | 45.4 | 277.8 | |
| LLaMA-8B-InstructRouter=true2025.10 | 41.2 | 309.1 | |
| LLaMA-3B-InstructRouter=true2025.10 | 38.6 | 413.3 | |
| LLaMA-8B-InstructRouter=false2025.10 | 38.4 | 320 | |
| LLaMA-3B-InstructRouter=false2025.10 | 35.2 | 422 | |
| Qwen-3B-InstructRouter=true2025.10 | 32.4 | 531.7 | |
| Qwen-3B-InstructRouter=false2025.10 | 30.2 | 536.3 | |
| LLaMA-8B-BaseRouter=true2025.10 | 20.2 | 531 | |
| LLaMA-8B-BaseRouter=false2025.10 | 17.2 | 536.7 | |
| LLaMA-3B-BaseRouter=true2025.10 | 15.8 | 544.3 | |
| LLaMA-3B-BaseRouter=false2025.10 | 11.8 | 548.4 |