Mathematical Reasoning on GSM8K (Accg, Accv, AUROC/APGR Analysis)
83AccgOLMo-7B† / Qwen-7B
Evaluation Results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| OLMo-7B† / Qwen-7BGenerator=OLMo-7B†, Verifier=Qwen-7B2026.03 | 83 | 90 | 7 | 76.3 | 53.8 | 74.8 | 80.8 | 96 | 65.7 | 95.6 | 95 | |
| Llama-3-8B† / Qwen-7BGenerator=Llama-3-8B†, Verifier=Qwen-7B2026.03 | 61 | 90 | 28 | 59.6 | 57.7 | 59 | 48 | 55.1 | 53.8 | 61.4 | 45.6 | |
| Mistral-7B† / Llama-3-8BGenerator=Mistral-7B†, Verifier=Llama-3-8B2026.03 | 51 | 61 | 11 | 65.5 | 53.9 | 71.9 | 68.7 | 87.2 | 63.8 | 99.8 | 88.7 | |
| Llama-1B / Qwen-7BGenerator=Llama-1B, Verifier=Qwen-7B2026.03 | 38 | 90 | 51 | 54.3 | 52.3 | 66.8 | 66.6 | 51.1 | 50.3 | 57.4 | 56.2 | |
| Llama-1B / Llama-3-8BGenerator=Llama-1B, Verifier=Llama-3-8B2026.03 | 38 | 61 | 23 | 54.3 | 52.3 | 65.5 | 60.6 | 52.1 | 50.5 | 66.2 | 58.5 | |
| Qwen-0.5B / Qwen-7BGenerator=Qwen-0.5B, Verifier=Qwen-7B2026.03 | 25 | 90 | 64 | 61.6 | 62.1 | 62.9 | 68.2 | 52 | 52.1 | 52.8 | 53.9 |