Mathematical Reasoning on MATH500, OlympiadBench, MinervaMath, GSM8K, and MMLU
58.1AccuracyRoBoN
Evaluation results
| Method | Links | |
|---|---|---|
| RoBoNn=2562025.12 | 58.1 | |
| RoBoNn=642025.12 | 57.6 | |
| BoNBackbone=qwen-coder-7b, n=2562025.12 | 56.4 | |
| BoNBackbone=llama-3.1-8b, n=2562025.12 | 56.3 | |
| BoN (Average across models)n=2562025.12 | 56.3 | |
| BoNBackbone=deepseek-coder-6.7b, n=2562025.12 | 56.2 | |
| BoNBackbone=qwen-math-7b, n=2562025.12 | 56.2 | |
| BoNBackbone=llama-3.1-8b, n=642025.12 | 56 | |
| BoNBackbone=qwen-coder-7b, n=642025.12 | 56 | |
| BoNBackbone=qwen-math-7b, n=642025.12 | 56 | |
| BoN (Average across models)n=642025.12 | 56 | |
| equaln=2562025.12 | 56 | |
| BoNBackbone=deepseek-coder-6.7b, n=642025.12 | 55.9 | |
| equaln=642025.12 | 55.2 | |
| BoNBackbone=qwen-math-7b, n=162025.12 | 54.9 | |
| BoNBackbone=llama-3.1-8b, n=162025.12 | 54.8 | |
| BoNBackbone=qwen-coder-7b, n=162025.12 | 54.8 | |
| BoN (Average across models)n=162025.12 | 54.8 | |
| RoBoNn=162025.12 | 54.8 | |
| BoNBackbone=deepseek-coder-6.7b, n=162025.12 | 54.6 | |
| equaln=162025.12 | 53.2 |