Adversarial Robustness Evaluation on AES
4ASRPAIR
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| PAIRGrader=Claude-Sonnet-4.62026.06 | 4 | -0.011 | 0.195 | |
| DANGrader=Claude-Sonnet-4.62026.06 | 9 | 0.004 | 0.589 | |
| PAIRGrader=GPT-5-mini2026.06 | 28 | 0.014 | 0.04 | |
| PAIRGrader=Gemini-3-flash2026.06 | 33.3 | 0.036 | 0.001 | |
| PAIRGrader=Qwen3-4B2026.06 | 38.3 | 0.051 | 0 | |
| PAIRGrader=Llama-3.1-8B2026.06 | 56 | 0.056 | 0 | |
| DANGrader=Qwen3-4B2026.06 | 59.6 | 0.107 | 0 | |
| MSJGrader=Claude-Sonnet-4.62026.06 | 72 | 0.152 | 0 | |
| MSJGrader=Llama-3.1-8B2026.06 | 79.5 | 0.237 | 0 | |
| MSJGrader=Gemini-3-flash2026.06 | 82 | 0.303 | 0 | |
| DANGrader=Gemini-3-flash2026.06 | 88.9 | 0.309 | 0 | |
| MSJGrader=GPT-5-mini2026.06 | 92 | 0.128 | 0 | |
| DANGrader=GPT-5-mini2026.06 | 99 | 0.427 | 0 | |
| MSJGrader=Qwen3-4B2026.06 | 100 | 0.36 | 0 | |
| DANGrader=Llama-3.1-8B2026.06 | 100 | 0.4 | 0 |