Mathematical Reasoning on AIME (Accuracy (mv %), Mean Questions, Mean Tokens)
86.7Accuracy (mv %)Qwen3-14B
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Qwen3-14BSetting=Adaptive, Mode=(t)2026.05 | 86.7 | 0 | 1,133 | |
| Qwen3-4BSetting=Adaptive, Mode=(t)2026.05 | 80 | 0 | 1,300 | |
| Qwen3-14BSetting=Baseline, Mode=(t)2026.05 | 80 | — | 11,871 | |
| Qwen3-4BSetting=Baseline, Mode=(t)2026.05 | 76.7 | — | 12,259 | |
| Olmo-3-7B-ThinkSetting=Baseline, Mode=(t)2026.05 | 76.7 | — | 15,429 | |
| Olmo-3-7B-ThinkSetting=Adaptive, Mode=(t)2026.05 | 76.7 | 1 | 33,243 | |
| Olmo-3-7B-ThinkSetting=Fixed, Mode=(t)2026.05 | 73.3 | 2 | 93,340 | |
| Qwen3-1.7BSetting=Adaptive, Mode=(t)2026.05 | 53.3 | 0.8 | 13,744 | |
| Qwen3-1.7BSetting=Baseline, Mode=(t)2026.05 | 48.3 | — | 13,570 | |
| Qwen3-14BSetting=Baseline, Mode=(nt)2026.05 | 43.3 | — | 4,324 | |
| Qwen3-14BSetting=Fixed, Mode=(nt)2026.05 | 40 | 3.1 | 17,730 | |
| Qwen3-4BSetting=Baseline, Mode=(nt)2026.05 | 36.7 | — | 1,271 | |
| Qwen3-14BSetting=Adaptive, Mode=(nt)2026.05 | 36.7 | 0.9 | 3,915 | |
| Qwen3-4BSetting=Adaptive, Mode=(nt)2026.05 | 30 | 1 | 5,317 | |
| Qwen3-4BSetting=Fixed, Mode=(nt)2026.05 | 28.6 | 2.3 | 20,523 | |
| Qwen3-1.7BSetting=Baseline, Mode=(nt)2026.05 | 16.7 | — | 2,933 | |
| Qwen3-1.7BSetting=Adaptive, Mode=(nt)2026.05 | 13.3 | 1.2 | 4,584 | |
| Qwen3-1.7BSetting=Fixed, Mode=(nt)2026.05 | 13.3 | 3.8 | 16,772 |