Language Understanding on MMLU (Acc, Peak, Dep)
66.5Accuracy (MMLU)CoT
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| CoTModel Series=Qwen2.5-7B, Budget setting=true2026.04 | 66.5 | 649 | 0.2 | |
| CoTModel Series=Llama3.1-8B, Budget setting=true2026.04 | 65.82 | 736 | 0.3 | |
| LThinker++Model Series=Qwen2.5-7B, Budget setting=true2026.04 | 60.27 | 1,133 | 3 | |
| VanillaModel Series=Llama3.1-8B, Budget setting=true2026.04 | 59.82 | 2,111 | 4.9 | |
| LThinker++Model Series=Llama3.1-8B, Budget setting=true2026.04 | 59.72 | 1,370 | 4.5 | |
| LThinker*Model Series=Llama3.1-8B, Budget setting=true2026.04 | 59.59 | 1,394 | 6.5 | |
| LThinker*Model Series=Qwen2.5-7B, Budget setting=true2026.04 | 57.68 | 1,141 | 4.1 | |
| VanillaModel Series=Qwen2.5-7B, Budget setting=true2026.04 | 57.58 | 2,432 | 6.5 | |
| TokenSkipModel Series=Llama3.1-8B, Budget setting=true2026.04 | 57.06 | 2,499 | 6.9 | |
| TokenSkipModel Series=Qwen2.5-7B, Budget setting=true2026.04 | 54.4 | 2,831 | 9.1 | |
| Distill-R1Model Series=Llama3.1-8B, Budget setting=true2026.04 | 53.46 | 582 | 0.8 | |
| Distill-R1Model Series=Qwen2.5-7B, Budget setting=true2026.04 | 51.7 | 2,483 | 7.5 |