Mathematical Reasoning on GSM8K (Score, SE, Delta(k=1))
95.6ScoreLlama3.1 70b Instruct
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Llama3.1 70b Instructn=1, 319, Decoding Strategy=Greedy2025.02 | 95.6 | 0.56 | 4.8 | |
| Llama3.1 70b Instructn=1, 319, Decoding Strategy=Random Sample, k=502025.02 | 95.3 | 0.45 | — | |
| Qwen 2.5 7B Instructn=1, 319, Decoding Strategy=Greedy2025.02 | 90.2 | 0.82 | 2.3 | |
| Qwen 2.5 7B Instructn=1, 319, Decoding Strategy=Random Sample, k=502025.02 | 90.2 | 0.65 | — | |
| Llama 3.1 8b Instructn=1, 319, Decoding Strategy=Greedy2025.02 | 86.1 | 0.95 | 18.6 | |
| Ministral 8B Instructn=1, 319, Decoding Strategy=Greedy2025.02 | 86.1 | 0.95 | 3.1 | |
| Llama 3.1 8b Instructn=1, 319, Decoding Strategy=Random Sample, k=502025.02 | 85.6 | 0.68 | — | |
| Ministral 8B Instructn=1, 319, Decoding Strategy=Random Sample, k=502025.02 | 84.9 | 0.73 | — |