Mathematical Reasoning on Math Benchmarks evaluated on Llama 3-70B
78.2GSM8K AccuracyBAdam
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| BAdamBackbone=Llama 3-70B2026.05 | 78.2 | 62.5 | 63.5 | 76.5 | 25.5 | 65.7 | 62 | |
| ChunkFTBackbone=Llama 3-70B2026.05 | 77.9 | 65.6 | 65.4 | 76.8 | 29.9 | 66.4 | 63.7 | |
| APOLLOBackbone=Llama 3-70B2026.05 | 75.6 | 63.6 | 63.7 | 73.4 | 28.6 | 64.8 | 61.6 | |
| LoRABackbone=Llama 3-70B2026.05 | 73.3 | 59.5 | 58.3 | 64.1 | 34.2 | 64.8 | 59 | |
| VanillaBackbone=Llama 3-70B2026.05 | 52.4 | 46.5 | 52.2 | 58.2 | 21.2 | 37.9 | 44.7 |