Image Pairs
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
100 image pairs Qwen2.5-VL-72B (test)
85.9Attack Success Rate (ASR)
8
100 image pairs Target: MiniCPM-o-4.5-8B (test)
90.2ASR
8
200 image pairs (test)
36.84PSNR
6
Loading the SOTA2 catalog…