Math Word Problem Solving on GSM8K (Acc, AUROC, ECE)
0.9636AccuracyQwen3 30B
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Qwen3 30Bconfidence_type=normalized, evaluation_protocol=self-evaluation2026.02 | 0.9636 | 0.787 | 0.026 | |
| Gemma3 12Bconfidence_type=normalized, evaluation_protocol=self-evaluation2026.02 | 0.9287 | 0.801 | 0.045 | |
| Qwen3 4Bconfidence_type=normalized, evaluation_protocol=self-evaluation2026.02 | 0.9249 | 0.773 | 0.069 | |
| Gemma3 4Bconfidence_type=normalized, evaluation_protocol=self-evaluation2026.02 | 0.8605 | 0.686 | 0.121 | |
| GLM4 9Bconfidence_type=normalized, evaluation_protocol=self-evaluation2026.02 | 0.6543 | 0.734 | 0.137 |