Image Captioning on Image Captioning Hard Criterion GPT-o1
53ASRCoTTA
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CoTTASurrogate=Ensemble2026.03 | 53 | 24.9 | |
| M-AttackSurrogate=Ensemble2026.03 | 29 | 10.5 | |
| FOA-AttackSurrogate=Ensemble2026.03 | 22 | 9.9 | |
| Agent-AttackSurrogate=Ensemble2026.03 | 7 | 3.3 | |
| AttackVLMSurrogate=B/162026.03 | 0 | 0 | |
| AttackVLMSurrogate=B/322026.03 | 0 | 0 | |
| AttackVLMSurrogate=Laion2026.03 | 0 | 0 | |
| AnyAttackSurrogate=Ensemble2026.03 | 0 | 0 |