Multilingual Safety Evaluation on 6 Safety Datasets High-Resource Languages
0.8659Safety Score (Fr)PG-Qwen-Smol
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| PG-Qwen-SmolModel Size=0.5B, Multilingual=true, Language Count=172025.12 | 0.8659 | 0.8596 | 0.8468 | 0.8528 | 0.8655 | |
| CREST-LARGEModel Size=0.5B, Multilingual=true, Language Count=1002025.12 | 0.8606 | 0.8608 | 0.8565 | 0.8533 | 0.8555 | |
| LlamaGuard3Model Size=8B, Multilingual=true, Language Count=82025.12 | 0.8407 | 0.8396 | 0.8278 | 0.8281 | 0.8345 | |
| CREST-BASEModel Size=0.25B, Multilingual=true, Language Count=1002025.12 | 0.8342 | 0.8299 | 0.8435 | 0.818 | 0.8321 | |
| DuoguardModel Size=0.5B, Multilingual=true, Language Count=292025.12 | 0.7381 | 0.6142 | 0.7686 | 0.674 | 0.7413 |