Full-response Safety Guardrail Classification on XSTest (test)
94.7F1 ScoreWildGuard-7B
Evaluation results
| Method | Links | |
|---|---|---|
| WildGuard-7B2026.06 | 94.7 | |
| Qwen3Guard-8B2026.06 | 92.3 | |
| SentGuardbackbone=Qwen3-4B-Instruct-25072026.06 | 91.2 | |
| LlamaGuard3-8B2026.06 | 90.4 | |
| Gemini-3.5-Flashprotocol=zero-shot2026.06 | 89.3 | |
| LlamaGuard4-12B2026.06 | 89 | |
| GPT-5.5protocol=zero-shot2026.06 | 88 |