Mathematical Reasoning on GSM8K (Accuracy & Retention Percentage)
102.5Retention PercentageLlada-Instruct
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Llada-InstructLayers Skipped=42026.03 | 102.5 | — | |
| Llada-InstructLayers Skipped=22026.03 | 101.3 | — | |
| Llada-InstructLayers Skipped=02026.03 | 100 | 83 | |
| Dream-BaseLayers Skipped=02026.03 | 100 | 77 | |
| Dream-InstructLayers Skipped=02026.03 | 100 | 81 | |
| Qwen2.5Layers Skipped=02026.03 | 100 | 49 | |
| Llada-InstructLayers Skipped=52026.03 | 97.7 | — | |
| Llada-InstructLayers Skipped=32026.03 | 96.7 | — | |
| Llada-InstructLayers Skipped=62026.03 | 91.8 | — | |
| Llada-InstructLayers Skipped=82026.03 | 91.8 | — | |
| Dream-InstructLayers Skipped=22026.03 | 84.6 | — | |
| Dream-BaseLayers Skipped=22026.03 | 76.8 | — | |
| Dream-InstructLayers Skipped=32026.03 | 71.2 | — | |
| Dream-BaseLayers Skipped=32026.03 | 62.6 | — | |
| Dream-BaseLayers Skipped=42026.03 | 44.5 | — | |
| Qwen2.5Layers Skipped=22026.03 | 34.9 | — | |
| Dream-InstructLayers Skipped=42026.03 | 25 | — | |
| Qwen2.5Layers Skipped=42026.03 | 20.6 | — | |
| Qwen2.5Layers Skipped=32026.03 | 15.9 | — |