Vision-Language Fuzzing on COCO
0Failure Rate (FR)Qwen2.5-VL-7B
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Qwen2.5-VL-7BModel Size=7B2026.02 | 0 | 100 | — | |
| Llama-3.2-11B-VisionModel Size=11B2026.02 | 1 | 12.84 | 93.98 | |
| Qwen2.5-VL-72BModel Size=72B2026.02 | 6.71 | 9.33 | 95.83 | |
| GPT-4o2026.02 | 7.59 | 4.5 | 91.37 | |
| FuzzingRLBase Model=Qwen2.5-VL-7B2026.02 | 34.47 | 7.75 | 91.5 |