Language Modeling on HellaSwag
58.5AccuracyTask Vector
Evaluation results
| Method | Links | |
|---|---|---|
| Task VectorBackbone=Llama2-7b, Runtime (min)=371.8672025.09 | 58.5 | |
| DOORBackbone=Llama2-7b, Runtime (min)=109.2302025.09 | 57.6 | |
| OriginalBackbone=Llama2-7b2025.09 | 57.2 | |
| NPOGDRBackbone=Llama2-7b, Runtime (min)=155.6332025.09 | 54.9 | |
| PRISMBackbone=Llama2-7b, Runtime (min)=40.1172025.09 | 54.6 | |
| SAM+NPOBackbone=Llama2-7b, Runtime (min)=75.5002025.09 | 54.4 | |
| GAGDRBackbone=Llama2-7b, Runtime (min)=122.0332025.09 | 53.5 | |
| NPOBackbone=Llama2-7b, Runtime (min)=215.9332025.09 | 52.8 | |
| GABackbone=Llama2-7b, Runtime (min)=205.6332025.09 | 49.7 |