Mathematical Reasoning on GSM8K (pass@4 Accuracy)
93.83pass@4 AccuracyPPCV
Evaluation Results
| Method | Links | |
|---|---|---|
| PPCVBackbone=Llama-3.1-8B-Instruct, k value=4, Number of samples per question=122026.02 | 93.83 | |
| Phi-DecodingBackbone=Llama-3.1-8B-Instruct, k value=4, Number of samples per question=122026.02 | 92.15 |