Jailbreaking on AdvBench (Classifier-Specific ASR@1)
0ASR@1 (No Refusal)ADV-LLM
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ADV-LLMVictim Model=GPT-oss-20B, Turn Type=Single-turn2026.02 | 0 | 0.4 | 0.8 | |
| Jigsaw PuzzleVictim Model=GPT-oss-20B, Turn Type=Multi-turn2026.02 | 10 | 0.8 | 3.7 | |
| Jailbreak-R1Victim Model=GPT-oss-20B, Turn Type=Single-turn2026.02 | 13.9 | 2.9 | 9.8 | |
| FITDVictim Model=GPT-oss-20B, Turn Type=Multi-turn2026.02 | 21.5 | 3.5 | 7.5 | |
| FlipAttackVictim Model=GPT-oss-20B, Turn Type=Single-turn2026.02 | 31 | 3.7 | 24.8 | |
| GOATVictim Model=GPT-oss-20B, Turn Type=Multi-turn2026.02 | 36.2 | 5.4 | 5.6 | |
| CoAVictim Model=GPT-oss-20B, Turn Type=Multi-turn2026.02 | 42.1 | 1.9 | 6.4 | |
| X-TeamingVictim Model=GPT-oss-20B, Turn Type=Multi-turn2026.02 | 45.6 | 15 | 30.2 | |
| CrescendoVictim Model=GPT-oss-20B, Turn Type=Multi-turn2026.02 | 58.5 | 21.2 | 40.2 | |
| SEMAVictim Model=GPT-oss-20B, Turn Type=Multi-turn2026.02 | 62.7 | 36 | 57.7 | |
| ActorAttackVictim Model=GPT-oss-20B, Turn Type=Multi-turn2026.02 | 88.3 | 6.5 | 19.2 |