Mathematical Reasoning on MATH (test) (Accuracy, Reasoning Length)
17.31AccuracyLTF
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| LTFBackbone=LLaMA-8B, Finetuning setting=True, Extreme compression=True2026.06 | 17.31 | 1 | |
| LTFBackbone=DS-1.5B, Finetuning setting=True, Extreme compression=True2026.06 | 17.2 | 1 | |
| LTFBackbone=LLaMA-3B, Finetuning setting=True, Extreme compression=True2026.06 | 16.18 | 1 | |
| ReGuLaRBackbone=DS-1.5B, Finetuning setting=True, Extreme compression=True2026.06 | 15.6 | 1 | |
| LTFBackbone=Average, Finetuning setting=True, Extreme compression=True2026.06 | 14.7 | 1 | |
| ReGuLaRBackbone=LLaMA-8B, Finetuning setting=True, Extreme compression=True2026.06 | 13.9 | 1 | |
| ReGuLaRBackbone=Average, Finetuning setting=True, Extreme compression=True2026.06 | 11.98 | 1 | |
| ReGuLaRBackbone=LLaMA-3B, Finetuning setting=True, Extreme compression=True2026.06 | 11.8 | 1 | |
| CoLaRBackbone=DS-1.5B, Finetuning setting=True, Extreme compression=True2026.06 | 11.13 | 62.2 | |
| CoLaRBackbone=LLaMA-8B, Finetuning setting=True, Extreme compression=True2026.06 | 9.13 | 67.19 | |
| CoconutBackbone=LLaMA-8B, Finetuning setting=True, Extreme compression=True2026.06 | 8.68 | 6 | |
| CoLaRBackbone=Average, Finetuning setting=True, Extreme compression=True2026.06 | 8.48 | 62.03 | |
| CoLaRBackbone=LLaMA-3B, Finetuning setting=True, Extreme compression=True2026.06 | 8.35 | 60.44 | |
| LTFBackbone=LLaMA-1B, Finetuning setting=True, Extreme compression=True2026.06 | 8.1 | 1 | |
| CoconutBackbone=LLaMA-3B, Finetuning setting=True, Extreme compression=True2026.06 | 7.72 | 6 | |
| CoconutBackbone=Average, Finetuning setting=True, Extreme compression=True2026.06 | 7.06 | 6 | |
| CoconutBackbone=DS-1.5B, Finetuning setting=True, Extreme compression=True2026.06 | 6.67 | 6 | |
| ReGuLaRBackbone=LLaMA-1B, Finetuning setting=True, Extreme compression=True2026.06 | 6.62 | 1 | |
| CoLaRBackbone=LLaMA-1B, Finetuning setting=True, Extreme compression=True2026.06 | 5.31 | 58.29 | |
| CoconutBackbone=LLaMA-1B, Finetuning setting=True, Extreme compression=True2026.06 | 5.18 | 6 |