Safety Evaluation on Anchor Safety Dataset
100Anchor AccuracyORIGINAL
Evaluation Results
| Method | Links | |
|---|---|---|
| ORIGINALModel Family=GEMMA-2-2B2026.03 | 100 | |
| CDAModel Family=GEMMA-2-2B2026.03 | 100 | |
| UGIDModel Family=GEMMA-2-2B2026.03 | 100 | |
| ORIGINALModel Family=QWEN2.5-3B2026.03 | 100 | |
| CDAModel Family=QWEN2.5-3B2026.03 | 100 | |
| UGIDModel Family=QWEN2.5-3B2026.03 | 100 | |
| ORIGINALModel Family=QWEN2.5-7B2026.03 | 100 | |
| CDAModel Family=QWEN2.5-7B2026.03 | 100 | |
| UGIDModel Family=QWEN2.5-7B2026.03 | 100 | |
| ORIGINALModel Family=QWEN2.5-14B2026.03 | 100 | |
| KLAADModel Family=QWEN2.5-14B2026.03 | 100 | |
| UGIDModel Family=QWEN2.5-14B2026.03 | 100 | |
| KLAADModel Family=GEMMA-2-2B2026.03 | 50 | |
| KLAADModel Family=QWEN2.5-3B2026.03 | 50 | |
| CDAModel Family=QWEN2.5-14B2026.03 | 50 | |
| KLAADModel Family=QWEN2.5-7B2026.03 | 0 |