Mathematical Reasoning on GSM8K (Accuracy, Rank)
75.2AccuracyBaseline
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| BaselineBackbone=LLaMA-3.1-8B-Instruct2026.05 | 75.2 | — | |
| SKOPBackbone=LLaMA-3.1-8B-Instruct2026.05 | 66.6 | 2.69 | |
| Angular SteerBackbone=LLaMA-3.1-8B-Instruct2026.05 | 60.5 | 4.75 | |
| SADIBackbone=LLaMA-3.1-8B-Instruct2026.05 | 58.6 | 3.38 | |
| CASTBackbone=LLaMA-3.1-8B-Instruct2026.05 | 41.8 | 5.63 | |
| LoRABackbone=LLaMA-3.1-8B-Instruct2026.05 | 29 | — | |
| DISCO-QBackbone=LLaMA-3.1-8B-Instruct2026.05 | 22.5 | 4.75 | |
| ITIBackbone=LLaMA-3.1-8B-Instruct2026.05 | 20.6 | 4.75 | |
| CAABackbone=LLaMA-3.1-8B-Instruct2026.05 | 14 | 5.44 | |
| Comm SteerBackbone=LLaMA-3.1-8B-Instruct2026.05 | 10.2 | 4.63 |