Knowledge Retention on MUSE Retain set (Dr)
56KnowMemGT
Evaluation results
| Method | Links | |
|---|---|---|
| GTBackbone Model=Llama-3.2 1B Instruct, Unlearning Ratio=10%2026.01 | 56 | |
| FullBackbone Model=Llama-3.2 1B Instruct, Unlearning Ratio=10%2026.01 | 55.5 | |
| NPO+Method Category=Training-based, Backbone Model=Llama-3.2 1B Instruct, Unlearning Ratio=10%2026.01 | 52.5 | |
| NPOMethod Category=Training-based, Backbone Model=Llama-3.2 1B Instruct, Unlearning Ratio=10%2026.01 | 52.1 | |
| PerTA-fisherMethod Category=Task Arithmetic-based, Backbone Model=Llama-3.2 1B Instruct, Unlearning Ratio=10%2026.01 | 46.4 | |
| PerTA-gradMethod Category=Task Arithmetic-based, Backbone Model=Llama-3.2 1B Instruct, Unlearning Ratio=10%2026.01 | 41.6 | |
| GDMethod Category=Training-based, Backbone Model=Llama-3.2 1B Instruct, Unlearning Ratio=10%2026.01 | 25.4 | |
| TVMethod Category=Task Arithmetic-based, Backbone Model=Llama-3.2 1B Instruct, Unlearning Ratio=10%2026.01 | 2.3 | |
| GAMethod Category=Training-based, Backbone Model=Llama-3.2 1B Instruct, Unlearning Ratio=10%2026.01 | 0.8 |