Risk Detection on EMBHAZARD (Held-out Set)
85Potential Risk ScoreHuman
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Human2026.05 | 85 | 92.8 | 58.8 | |
| EMBGUARD-4BParameters=4B2026.05 | 77 | 60.7 | 53.3 | |
| Qwen-3-VL-235B-a22bParameters=235B2026.05 | 70.3 | 71.7 | 46.2 | |
| GPT-5.12026.05 | 67.3 | 67.2 | 61.9 | |
| EMBGUARD-2BParameters=2B2026.05 | 61 | 39.2 | 22.7 |