Knowledge Retention on MUSE utility
18.4Util ScoreLUNAR
Evaluation Results
| Method | Links | |
|---|---|---|
| LUNARBackbone=Llama-3.1-8B-Instruct2026.05 | 18.4 | |
| LUNARBackbone=Qwen-3-8B2026.05 | 15.9 | |
| Surgical GABackbone=Qwen-3-8B2026.05 | 2 | |
| MANSUBackbone=Qwen-3-8B2026.05 | 1.9 | |
| NPOBackbone=Llama-3.1-8B-Instruct2026.05 | 1.3 | |
| NPOBackbone=Qwen-3-8B2026.05 | 1.1 | |
| Global GABackbone=Qwen-3-8B2026.05 | 0.7 | |
| MANSUBackbone=Llama-3.1-8B-Instruct2026.05 | 0.6 | |
| GU+SimNPOBackbone=Qwen-3-8B2026.05 | 0.5 | |
| Global GABackbone=Llama-3.1-8B-Instruct2026.05 | 0 | |
| Surgical GABackbone=Llama-3.1-8B-Instruct2026.05 | 0 | |
| SimNPOBackbone=Llama-3.1-8B-Instruct2026.05 | 0 | |
| GU+SimNPOBackbone=Llama-3.1-8B-Instruct2026.05 | 0 | |
| SimNPOBackbone=Qwen-3-8B2026.05 | 0 |