Mathematical Reasoning on GSM8K (test) (Accuracy, Average response length)
95.7AccuracyRPAM
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| RPAMMethod Category=Data-dependent Merging2026.01 | 95.7 | 1,086 | |
| CoDMethod Category=Prompt-guided2026.01 | 95.6 | 1,017 | |
| Average MergingMethod Category=Data-free Merging2026.01 | 95.4 | 1,088 | |
| Ada-R1Method Category=Training-based2026.01 | 95.3 | 1,161 | |
| Qwen3-4B-ThinkingMethod Category=Base Model, Thinking Strategy=Long-CoT2026.01 | 95.2 | 1,521 | |
| Task ArithmeticMethod Category=Data-free Merging2026.01 | 95.2 | 1,181 | |
| DARE-LinearMethod Category=Data-free Merging2026.01 | 95.1 | 1,641 | |
| AIMMethod Category=Data-dependent Merging2026.01 | 95.1 | 1,060 | |
| TIES MergingMethod Category=Data-free Merging2026.01 | 94.8 | 1,170 | |
| ACMMethod Category=Data-dependent Merging2026.01 | 94.7 | 925 | |
| Qwen3-4B-InstructMethod Category=Base Model, Thinking Strategy=Short-CoT2026.01 | 93 | 374 |