LLM Unlearning on MT-Bench
5.62FluencyBase Model
Evaluation results
| Method | Links | |
|---|---|---|
| Base ModelBackbone=Llama-3-8B2025.02 | 5.62 | |
| RMUBackbone=Llama-3-8B2025.02 | 5.39 | |
| NPOGDRBackbone=Llama-3-8B2025.02 | 5.18 | |
| GAGDRBackbone=Llama-3-8B2025.02 | 3.97 | |
| Base ModelBackbone=Mistral-7B2025.02 | 1.71 | |
| RMUBackbone=Mistral-7B2025.02 | 1.58 | |
| NPOGDRBackbone=Mistral-7B2025.02 | 1.04 | |
| NPOKLRBackbone=Llama-3-8B2025.02 | 1.03 | |
| GAKLRBackbone=Llama-3-8B2025.02 | 1.01 | |
| TVBackbone=Llama-3-8B2025.02 | 1.01 | |
| GABackbone=Llama-3-8B2025.02 | 1 | |
| NPOBackbone=Llama-3-8B2025.02 | 1 | |
| GABackbone=Mistral-7B2025.02 | 1 | |
| GAGDRBackbone=Mistral-7B2025.02 | 1 | |
| GAKLRBackbone=Mistral-7B2025.02 | 1 | |
| NPOBackbone=Mistral-7B2025.02 | 1 | |
| NPOKLRBackbone=Mistral-7B2025.02 | 1 | |
| TVBackbone=Mistral-7B2025.02 | 1 |