Attack Success Rate (ASR@N) on AdvBench
37.6ASR@1SEMA
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| SEMAVictim=GPT-oss-20B2026.02 | 37.6 | 68.7 | 76.3 | 80.8 | 81.8 | 82.5 | 83.5 | 84.2 |
Loading the SOTA2 catalog…
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| SEMAVictim=GPT-oss-20B2026.02 | 37.6 | 68.7 | 76.3 | 80.8 | 81.8 | 82.5 | 83.5 | 84.2 |