LLM Attack Effectiveness on DeepSeek-R1-Distill-Llama-8B serving environment
0.08TTFT (s)Benign
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Benignmode=No Attack2026.02 | 0.08 | 0.14 | 37.51 | 41.15 | 0 | 0 | — | |
| Engorgio2026.02 | 0.1 | 0.14 | 48.86 | 55.8 | 0 | 2,027 | 1.17 | |
| LoopLLM2026.02 | 1.07 | 0.56 | 57.37 | 115.3 | 89 | 992 | 1.06 | |
| F&Sinput=plain-text2026.02 | 5.31 | 110.27 | 89.52 | 478.9 | 1,098 | 321 | 0.95 | |
| ExtendAttack2026.02 | 13.29 | 220.58 | 188.86 | 1,021.82 | 773 | 168 | 0.85 | |
| F&Sinput=ExtendAttack2026.02 | 16.08 | 284.41 | 157.8 | 1,215.48 | 506 | 183 | 0.86 |