Mathematical Reasoning on GSM8K (Accuracy & Avg gain)
94.92AccuracyMIND
Evaluation results
| Method | Links | ||
|---|---|---|---|
| MINDBackbone=Qwen2.5-7B, Fusion=with2026.01 | 94.92 | — | |
| SbSBackbone=Qwen2.5-7B, Fusion=none2026.01 | 94.77 | 2.23 | |
| EDITBackbone=Qwen2.5-7B, Fusion=none2026.01 | 94.28 | 1.53 | |
| MoDEBackbone=Qwen2.5-7B, Fusion=none2026.01 | 94.16 | 2.23 | |
| BaseBackbone=Qwen2.5-7B-Instruct, Fusion=none2026.01 | 92.36 | 3.99 | |
| MINDBackbone=Qwen2.5-7B, Fusion=without2026.01 | 92 | 2.31 | |
| MCCBackbone=Qwen2.5-7B, Fusion=none2026.01 | 90.52 | 2.71 | |
| MINDBackbone=Llama3.1-8B, Fusion=with2026.01 | 87.51 | — | |
| SbSBackbone=Llama3.1-8B, Fusion=none2026.01 | 86.96 | 0.67 | |
| EDITBackbone=Llama3.1-8B, Fusion=none2026.01 | 86.91 | 0.62 | |
| MCCBackbone=Llama3.1-8B, Fusion=none2026.01 | 85.01 | 2.63 | |
| MoDEBackbone=Llama3.1-8B, Fusion=none2026.01 | 84 | 4.28 | |
| MINDBackbone=Llama3.1-8B, Fusion=without2026.01 | 83.96 | 3.78 | |
| BaseBackbone=Llama3.1-8B-Instruct, Fusion=none2026.01 | 83.89 | 4.11 | |
| MINDBackbone=Qwen2.5-1.5B, Fusion=with2026.01 | 76.5 | — | |
| EDITBackbone=Qwen2.5-1.5B, Fusion=none2026.01 | 74.5 | 2.71 | |
| SbSBackbone=Qwen2.5-1.5B, Fusion=none2026.01 | 73.91 | 3.73 | |
| BaseBackbone=Qwen2.5-1.5B-Instruct, Fusion=none2026.01 | 72.81 | 3.63 | |
| MCCBackbone=Qwen2.5-1.5B, Fusion=none2026.01 | 70.54 | 3.16 | |
| MINDBackbone=Qwen2.5-1.5B, Fusion=without2026.01 | 70.52 | 4.41 | |
| MoDEBackbone=Qwen2.5-1.5B, Fusion=none2026.01 | 70.31 | 3.72 |