GPT-4o
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Jailbreak Attack | GPT-4o API | ASR84 | 16 | |
| Targeted Adversarial Attack | GPT-4o | ASR860 | 12 | |
| Open-ended image description | GPT-4o assisted evaluation | Accuracy8.76 | 8 | |
| Detection of paraphrased text | GPT-4o-mini Paraphrased | ROC AUC (FPR=1%)0.4231 | 8 | |
| Jailbreak Attack | GPT-4o (test) | ASR95 | 6 | |
| Denial-of-Service Attack | GPT-4o-mini 2024-07-18 (test) | Response Length16,384 | 6 | |
| Jailbreak | GPT-4o 29 May 2025 release | ASR98.46 | 5 | |
| Policy Corruption Evaluation | GPT-4o mini | Compliance Score3.53 | 5 | |
| Keyword Matching Attack | GPT-4o | KMR (alpha)73 | 4 | |
| Sycophancy-Induced Spiral Dynamics Intervention | GPT-4o high-sycophancy deployment (n = 200, T = 30) | Spiral Rate16.5 | 3 | |
| Jailbreaking | GPT-4o efficiency analysis | Attack Success Rate (ASR)65.7 | 3 |
Showing 11 of 11 rows