Mathematical Reasoning on AIME25 (Accuracy, Average response length)
80.3AccuracyQwen3-4B-Thinking
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Qwen3-4B-ThinkingMethod Category=Base Model, Thinking Strategy=Long-CoT2026.01 | 80.3 | 23,912 | |
| CoDMethod Category=Prompt-guided2026.01 | 76.7 | 17,338 | |
| Ada-R1Method Category=Training-based2026.01 | 68.9 | 11,969 | |
| Task ArithmeticMethod Category=Data-free Merging2026.01 | 67.8 | 11,395 | |
| RPAMMethod Category=Data-dependent Merging2026.01 | 67.8 | 10,157 | |
| TIES MergingMethod Category=Data-free Merging2026.01 | 60 | 10,891 | |
| AIMMethod Category=Data-dependent Merging2026.01 | 60 | 9,934 | |
| Average MergingMethod Category=Data-free Merging2026.01 | 57.8 | 10,099 | |
| DARE-LinearMethod Category=Data-free Merging2026.01 | 56.7 | 12,247 | |
| ACMMethod Category=Data-dependent Merging2026.01 | 54.4 | 11,080 | |
| Qwen3-4B-InstructMethod Category=Base Model, Thinking Strategy=Short-CoT2026.01 | 50 | 7,368 |