Vision-text safety classification on MSSBench Embodied
61.97AUPRC (Prompt)SEA-Guard-4B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SEA-Guard-4BZero-shot=true2026.02 | 61.97 | 59.71 | |
| SEA-Guard-8BZero-shot=true2026.02 | 57.43 | 60.97 | |
| SEA-Guard-12BZero-shot=true2026.02 | 53.1 | 59.61 | |
| Gemma3-12B-ITZero-shot=true2026.02 | 51 | 53.94 | |
| Gemma3-4B-ITZero-shot=true2026.02 | 50.79 | 54.13 | |
| Qwen3-VL-4B-ITZero-shot=true2026.02 | 50.66 | 58.58 | |
| SEA-LION-v4-Qwen-VLZero-shot=true2026.02 | 50.33 | 57.24 | |
| SEA-LION-v4-Qwen-VLZero-shot=true, variant=v22026.02 | 50.17 | 55.62 | |
| Qwen3-VL-8B-ITZero-shot=true2026.02 | 50 | 59.41 |