Mathematical Reasoning and Coding on Polaris
48.79Peak Accuracy@8balanced-agg
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| balanced-aggModel=Qwen3-1.7B, Aggregation=balanced-agg2026.04 | 48.79 | 60.94 | 46.4 | 55.5 | |
| token-aggModel=Qwen3-1.7B, Aggregation=token-agg2026.04 | 48.42 | 60.53 | 43.49 | 55.57 | |
| seq-aggModel=Qwen3-1.7B, Aggregation=seq-agg2026.04 | 48.12 | 59.5 | 46.14 | 57.3 | |
| token-aggModel=Qwen2.5-Math-7B, Aggregation=token-agg2026.04 | 35.39 | 47.68 | 32.92 | 45.27 | |
| balanced-aggModel=Qwen2.5-Math-7B, Aggregation=balanced-agg2026.04 | 34.23 | 47.5 | 33.19 | 44.43 | |
| seq-aggModel=Qwen2.5-Math-7B, Aggregation=seq-agg2026.04 | 34.05 | 46.97 | 31.72 | 43.6 |