Risk detection on NuInstruct (test)
2.01MAEOmniSpace
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| OmniSpaceEvaluation Protocol=Fine-tuned2026.06 | 2.01 | 64.81 | 40.36 | 90.66 | 48.45 | |
| BaselineEvaluation Protocol=Fine-tuned2026.06 | 2.47 | 59.51 | 35.32 | 86.71 | 44.81 | |
| VGGDriveEvaluation Protocol=Fine-tuned2026.06 | 3.08 | 56.37 | 37.49 | 81.13 | 42.98 | |
| InMLLMEvaluation Protocol=Fine-tuned2026.06 | 9.08 | 32.48 | 21.93 | 35.2 | 20.13 | |
| GPT-4oEvaluation Protocol=Zero-shot2026.06 | 9.93 | 10.64 | 0 | 7.08 | 1.95 | |
| RoboTronEvaluation Protocol=Zero-shot2026.06 | 19.36 | 2.57 | 0 | 8.06 | 0 | |
| Qwen2.5-VLEvaluation Protocol=Zero-shot2026.06 | 24.1 | 0.63 | 0 | 5.56 | 0 | |
| LLAVA-OVEvaluation Protocol=Zero-shot2026.06 | 87.04 | 3.75 | 0 | 8.55 | 0 |