Instruction Following and Safety Evaluation on Combined Alpaca, HarmfulQA, Jailbreak (test)
2.77Helpful ScoreQwen2.5-14B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen2.5-14BModel backbone=Qwen2.5-14B, Training status=Base2025.10 | 2.77 | 14 | |
| Boundary GuidanceModel backbone=Qwen2.5-7B, Training status=Fine-tuned (FT)2025.10 | 2.75 | 4 | |
| Qwen2.5-7BModel backbone=Qwen2.5-7B, Training status=Base2025.10 | 2.72 | 19 | |
| Boundary GuidanceModel backbone=Qwen2.5-14B, Training status=Fine-tuned (FT)2025.10 | 2.71 | 3 | |
| Boundary GuidanceModel backbone=Gemma-2-9B-it, Training status=Fine-tuned (FT)2025.10 | 2.66 | 4 | |
| Gemma-2-9B-itModel backbone=Gemma-2-9B-it, Training status=Base2025.10 | 2.63 | 7 | |
| Boundary GuidanceModel backbone=Qwen2.5-0.5B, Training status=Fine-tuned (FT)2025.10 | 2.26 | 17 | |
| Qwen2.5-0.5BModel backbone=Qwen2.5-0.5B, Training status=Base2025.10 | 2.13 | 27 |