Instruction-following on Average (Alpaca, Samsum, ChatDoctor)
0.508BRT ScoreSAP
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SAPBackbone=Llama2-7B2025.05 | 0.508 | 4.64 | 21.7 | |
| SFTBackbone=Llama2-7B2025.05 | 0.506 | 4.67 | 31.93 | |
| SafeInstrBackbone=Llama2-7B2025.05 | 0.504 | 4.67 | 26.03 | |
| SaLoRABackbone=Llama2-7B2025.05 | 0.501 | 4.72 | 28.03 | |
| BoosterBackbone=Llama2-7B2025.05 | 0.497 | 4.72 | 26.2 | |
| SAFTBackbone=Llama2-7B2025.05 | 0.497 | 4.79 | 29.77 | |
| LisaBackbone=Llama2-7B2025.05 | 0.495 | 4.78 | 25.53 | |
| VaccineBackbone=Llama2-7B2025.05 | 0.491 | 4.76 | 27.27 | |
| Base modelBackbone=Llama2-7B2025.05 | 0.416 | 13.08 | 30.9 |