Mathematical Reasoning on SAT Math, AMC23, AIME24, and Olympiad Bench Average (out-domain)
65.77Average AccuracyQwen3-4B
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Qwen3-4BBase Model=Qwen3-4B2026.02 | 65.77 | 2,472.8 | 0 | 0 | |
| ThinkPrune-500Base Model=Qwen3-4B2026.02 | 64.96 | 2,109.8 | 0.09 | 0.02 | |
| ThinkPrune-1kBase Model=Qwen3-4B2026.02 | 64.46 | 2,006.9 | 0.09 | -0.01 | |
| ACPOBase Model=Qwen3-4B2026.02 | 63.12 | 1,675.6 | 0.12 | -0.08 | |
| CRTBase Model=Qwen3-4B2026.02 | 62.99 | 1,361.4 | 0.24 | 0.03 | |
| O1-Pruner-4.0Base Model=Qwen3-4B2026.02 | 62.66 | 1,394.3 | 0.2 | -0.04 | |
| ShorterBetterBase Model=Qwen3-4B2026.02 | 61.22 | 1,343.4 | 0.11 | -0.23 |