In-Context Learning on Llama3-8B Scenario 3: More layers (L={0,4,8,...})
80.43AccuracyLTV
Evaluation Results
| Method | Links | |
|---|---|---|
| LTVModel=Llama3-8B, Layer={0,4,8,...}2025.09 | 80.43 | |
| Vanilla TVModel=Llama3-8B, Layer={0,4,8,...}2025.09 | 37.05 | |
| FVModel=Llama3-8B, Layer={0,4,8,...}2025.09 | 4.69 |