Mathematical Reasoning on MATH (Retention)
28.9RetentionRetainOnlyFT
Evaluation results
| Method | Links | |
|---|---|---|
| RetainOnlyFTModel=Llama-3.2-3B-Instruct2026.04 | 28.9 | |
| EUModel=Llama-3.2-3B-Instruct, Hyperparameter (λ)=0.62026.04 | 28.66 | |
| Llama-3.2-3B-InstructModel=Llama-3.2-3B-Instruct2026.04 | 22.96 | |
| EraserModel=Llama-3.2-3B-Instruct, Hyperparameter (λ)=0.62026.04 | 22.18 | |
| DPOModel=Llama-3.2-3B-Instruct2026.04 | 22.14 | |
| SKUModel=Llama-3.2-3B-Instruct2026.04 | 21.14 | |
| UnlearningModel=Llama-3.2-3B-Instruct, Hyperparameter (λ)=0.62026.04 | 19.5 | |
| EraserModel=OLMo-2-7B-Instruct, Hyperparameter (λ)=0.62026.04 | 19.3 | |
| DPOModel=Llama-3.2-1B-Instruct2026.04 | 18.36 | |
| EraserModel=Llama-3.2-1B-Instruct, Hyperparameter (λ)=0.62026.04 | 15.92 | |
| RetainOnlyFTModel=Llama-3.2-1B-Instruct2026.04 | 15.56 | |
| UnlearningModel=Llama-3.2-1B-Instruct, Hyperparameter (λ)=0.62026.04 | 15.42 | |
| EUModel=Llama-3.2-1B-Instruct, Hyperparameter (λ)=0.62026.04 | 15.06 | |
| Llama-3.2-1B-InstructModel=Llama-3.2-1B-Instruct2026.04 | 14.88 | |
| EUModel=OLMo-2-7B-Instruct, Hyperparameter (λ)=0.62026.04 | 14.44 | |
| SKUModel=Llama-3.2-1B-Instruct2026.04 | 12.9 | |
| EraserModel=OLMo-2-1B-Instruct, Hyperparameter (λ)=0.62026.04 | 12.34 | |
| UnlearningModel=OLMo-2-1B-Instruct, Hyperparameter (λ)=0.62026.04 | 11.48 | |
| EUModel=OLMo-2-1B-Instruct, Hyperparameter (λ)=0.62026.04 | 11.24 | |
| RetainOnlyFTModel=OLMo-2-7B-Instruct2026.04 | 11.1 | |
| RetainOnlyFTModel=OLMo-2-1B-Instruct2026.04 | 11.06 | |
| DPOModel=OLMo-2-1B-Instruct2026.04 | 10.54 | |
| UnlearningModel=OLMo-2-7B-Instruct, Hyperparameter (λ)=0.62026.04 | 10.32 | |
| OLMo-2-1B-InstructModel=OLMo-2-1B-Instruct2026.04 | 8.66 | |
| DPOModel=OLMo-2-7B-Instruct2026.04 | 8.26 | |
| OLMo-2-7B-InstructModel=OLMo-2-7B-Instruct2026.04 | 6.86 | |
| SKUModel=OLMo-2-1B-Instruct2026.04 | 5.94 | |
| SKUModel=OLMo-2-7B-Instruct2026.04 | 5.38 |