Mathematical Reasoning on GSM8K (test) (Accuracy, AUC_wrong (S), Spearman, AUC_wrong (S50))
0.749AccuracyLlama-3.1-8B-Inst
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Llama-3.1-8B-InstN=1319, Decoding=stochastic2026.02 | 0.749 | 0.692 | -0.288 | 0.598 | |
| Llama-3.2-3B-InstN=1319, Decoding=stochastic2026.02 | 0.707 | 0.687 | -0.295 | 0.618 | |
| Qwen2.5-7B-InstN=1319, Decoding=stochastic2026.02 | 0.5 | 0.644 | -0.25 | 0.581 | |
| Llama-3.2-1B-InstN=1319, Decoding=stochastic2026.02 | 0.434 | 0.673 | -0.298 | 0.615 | |
| Qwen2.5-1.5B-InstN=1319, Decoding=stochastic2026.02 | 0.416 | 0.643 | -0.244 | 0.565 | |
| Qwen2.5-0.5B-InstN=1319, Decoding=stochastic2026.02 | 0.301 | 0.714 | -0.341 | 0.654 |