Mathematical Reasoning Accuracy on GSM8K (test)
83.55AccuracyBBoxER
Evaluation Results
| Method | Links | |
|---|---|---|
| BBoxERBase Model=Qwen-2.5-3B-instruct, Budget (b)=300, 8-shot evaluation=true, Optimization Algorithm=D-CMA, Update Type=normalization layers2025.07 | 83.55 | |
| BBoxERBase Model=Qwen-2.5-3B-instruct, Budget (b)=150, 8-shot evaluation=true, Optimization Algorithm=D-CMA, Update Type=normalization layers2025.07 | 82.9 | |
| Qwen-2.5-3B-instruct8-shot evaluation=true2025.07 | 79.98 | |
| BBoxERBase Model=Llama3.1-8B-Instruct, Budget (b)=300, 8-shot evaluation=true, Optimization Algorithm=D-CMA, Update Type=normalization layers2025.07 | 78.57 | |
| BBoxERBase Model=Llama3.1-8B-Instruct, Budget (b)=150, 8-shot evaluation=true, Optimization Algorithm=D-CMA, Update Type=normalization layers2025.07 | 77.79 | |
| Llama3.1-8B-Instruct8-shot evaluation=true2025.07 | 77.26 | |
| BBoxERBase Model=Llama3.1-8B, Budget (b)=300, 8-shot evaluation=true, Optimization Algorithm=D-CMA, Update Type=normalization layers2025.07 | 56.65 | |
| BBoxERBase Model=Llama3.1-8B, Budget (b)=150, 8-shot evaluation=true, Optimization Algorithm=D-CMA, Update Type=normalization layers2025.07 | 55.24 | |
| Llama3.1-8B8-shot evaluation=true2025.07 | 54.97 |