Binary safety classification on ML-BENCH (test)
97F1 (Seed Query)ML-GUARD-7B
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| ML-GUARD-7BModel Size=7B2026.05 | 97 | 90 | 61 | 99 | 97 | 44 | 6 | 19 | 0 | 92 | |
| gpt-oss-safeguard-20BModel Size=20B2026.05 | 85 | 84 | 61 | 94 | 72 | 44 | 27 | 0 | 0 | 44 | |
| ML-GUARD-1.5BModel Size=1.5B2026.05 | 85 | 70 | 35 | 83 | 57 | 21 | 13 | 5 | 0 | 50 | |
| PolyGuard-Qwen2026.05 | 72 | 34 | 33 | 68 | 21 | 24 | 22 | 2 | 17 | 7 | |
| Nemotron-8BModel Size=8B2026.05 | 62 | 23 | 17 | 59 | 13 | 9 | 16 | 0 | 0 | 18 | |
| Qwen3Guard-Gen-8BModel Size=8B2026.05 | 61 | 8 | 6 | 44 | 4 | 3 | 6 | 0 | 0 | 2 | |
| Qwen3Guard-Gen-4BModel Size=4B2026.05 | 59 | 6 | 4 | 40 | 3 | 2 | 5 | 0 | 0 | 1 | |
| Qwen3Guard-Gen-0.6BModel Size=0.6B2026.05 | 57 | 9 | 8 | 48 | 5 | 4 | 12 | 2 | 0 | 6 | |
| Llama-Guard-3-8BModel Size=8B2026.05 | 49 | 6 | 10 | 33 | 3 | 5 | 3 | 1 | 0 | 3 | |
| Llama-Guard-3-1BModel Size=1B2026.05 | 48 | 47 | 45 | 46 | 47 | 45 | 46 | 45 | 45 | 45 | |
| DuoGuard-1.5BModel Size=1.5B2026.05 | 46 | 32 | 4 | 36 | 21 | 2 | 20 | 10 | 11 | 20 | |
| Llama-Guard-4-12BModel Size=12B2026.05 | 45 | 27 | 5 | 30 | 16 | 3 | 4 | 1 | 13 | 18 | |
| Omni-moderation2026.05 | 2 | 0 | 2 | 1 | 0 | 1 | 0 | 0 | 0 | 0 |