Safety Classification on Combined Chinese Safety Datasets
77.93Overall Average ScoreCHILLGuard-4B
Evaluation Results
| Method | Links | |
|---|---|---|
| CHILLGuard-4BScale=4∼7B2026.06 | 77.93 | |
| CHILLGuard-8BScale=8B+2026.06 | 77.75 | |
| CHILLGuard-1.7BScale=0∼3B2026.06 | 74.88 | |
| PolyGuard-7BScale=4∼7B2026.06 | 69.76 | |
| Qwen3Guard-0.6B-StrictScale=0∼3B2026.06 | 67.8 | |
| LlamaGuard3-8BScale=8B+2026.06 | 65.04 | |
| Qwen3Guard-0.6B-LooseScale=0∼3B2026.06 | 63.06 | |
| WildGuard-7BScale=4∼7B2026.06 | 62.46 | |
| Qwen3Guard-4B-LooseScale=4∼7B2026.06 | 62.22 | |
| Qwen3Guard-8B-StrictScale=8B+2026.06 | 61.5 | |
| Qwen3Guard-4B-StrictScale=4∼7B2026.06 | 61.25 | |
| Qwen3Guard-8B-LooseScale=8B+2026.06 | 52.84 | |
| ShieldGemma-9BScale=8B+2026.06 | 51.2 | |
| LlamaGuard4-12BScale=8B+2026.06 | 50.95 | |
| NemoGuard-8BScale=8B+2026.06 | 49.74 | |
| ShieldGemma-27BScale=8B+2026.06 | 47.19 |