Machine Unlearning on Dolly-15k Mistral-7B variant (Seen)
86Seen ASRRetain Only
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Retain OnlyBackbone=Llama-3-8B-Instruct, PEFT Framework=LoRA (r=16, alpha=16, dropout=0.05)2025.12 | 86 | 6 | |
| Base (Poisoned)Backbone=Llama-3-8B-Instruct2025.12 | 84.4 | 5 | |
| Fisher UnlearnBackbone=Llama-3-8B-Instruct, PEFT Framework=LoRA (r=16, alpha=16, dropout=0.05)2025.12 | 83 | 3.67 | |
| Fisher UnlearnBackbone=Mistral-7B, Unlearning Category=Approximate unlearning methods2025.12 | 66.9 | 5.67 | |
| Retain OnlyBackbone=Mistral-7B, Unlearning Category=Approximate unlearning methods2025.12 | 66.5 | 4.67 | |
| Base (Poisoned)Backbone=Mistral-7B, Unlearning Category=Approximate unlearning methods2025.12 | 62.5 | 3 | |
| LoReUnBackbone=Mistral-7B, Unlearning Category=Approximate unlearning methods2025.12 | 41.4 | 3 | |
| GA UnlearnBackbone=Mistral-7B, Unlearning Category=Approximate unlearning methods2025.12 | 38.4 | 3.33 | |
| GA UnlearnBackbone=Llama-3-8B-Instruct, PEFT Framework=LoRA (r=16, alpha=16, dropout=0.05)2025.12 | 25.3 | 3.33 | |
| RapidUnBackbone=Mistral-7B, Unlearning Category=Approximate unlearning methods2025.12 | 22.4 | 1.33 | |
| LoReUnBackbone=Llama-3-8B-Instruct, PEFT Framework=LoRA (r=16, alpha=16, dropout=0.05)2025.12 | 21.4 | 2 | |
| RapidUnBackbone=Llama-3-8B-Instruct, PEFT Framework=LoRA (r=16, alpha=16, dropout=0.05)2025.12 | 15.3 | 1 | |
| Retrain (reference)Backbone=Llama-3-8B-Instruct2025.12 | 4.97 | — | |
| Retrain (reference)Backbone=Mistral-7B, Unlearning Category=Reference2025.12 | 3 | — |