Mathematical Reasoning on LiveBench
53.6AccuracyT-MLP
Evaluation Results
| Method | Links | |
|---|---|---|
| T-MLPBackbone=LLaMA-3.1-8B-Instruct2026.05 | 53.6 | |
| T-AllBackbone=LLaMA-3.1-8B-Instruct2026.05 | 52.3 | |
| BaseBackbone=LLaMA-3.1-8B-Instruct2026.05 | 47.9 | |
| Vanilla SFTBackbone=LLaMA-3.1-8B-Instruct2026.05 | 47.9 | |
| Vanilla SFTBackbone=LLaMA-3.2-3B-Instruct2026.05 | 45.4 | |
| T-MLPBackbone=LLaMA-3.2-3B-Instruct2026.05 | 44.6 | |
| T-AllBackbone=LLaMA-3.2-3B-Instruct2026.05 | 42.4 | |
| BaseBackbone=Mistral-7B-Instruct2026.05 | 39.7 | |
| T-MLPBackbone=Mistral-7B-Instruct2026.05 | 38 | |
| T-AllBackbone=Mistral-7B-Instruct2026.05 | 37.6 | |
| BaseBackbone=LLaMA-3.2-3B-Instruct2026.05 | 35.6 | |
| Vanilla SFTBackbone=Mistral-7B-Instruct2026.05 | 31 |