DALL-E
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Membership Inference Attack | DALL·E (test) | TPR @ 5% FPR29.4 | 54 | |
| Image Copy Detection | DALL-E 2 (test) | Average Similarity0.826 | 28 | |
| Membership Inference Attack | DALL-E | Accuracy72.3 | 26 | |
| Generated Image Detection | DALL-E 3 | AP98.1 | 15 | |
| Synthetic Image Detection | DALL-E 2 (full) | Acc@EER77.3 | 12 | |
| Fake image detection | DALL·E | AA98.81 | 9 | |
| Synthetic Image Detection | DALL-E 2 (Academic) | AUCROC0.727 | 9 | |
| Synthetic Image Detection | DALL-E 3 (proprietary) | AUCROC0.759 | 9 | |
| Synthetic Image Detection | DALL-E 2 (proprietary) | AUCROC85.4 | 9 | |
| Synthetic image detection | DALL-E | Accuracy98.8 | 9 | |
| Nudity Jailbreaking Transfer | DALL·E Universal nudity jailbreaking prompts 3 | Transfer Success Rate35.58 | 7 | |
| Image Watermarking | DALL-E 3 | PSNR18.28 | 7 | |
| Watermark Robustness | DALL-E 3 | Robustness: Brightness100 | 7 | |
| Adversarial Attack | DALL·E 3 commercial (test) | BR0.95 | 7 | |
| Jailbreak Attack | DALL·E 3 | TASR (%)12.98 | 6 | |
| Synthetic Image Detection | DALL-E 3 (full) | Acc@EER69.8 | 6 | |
| Synthetic Image Detection | DALL-E Anime 2 | AP62 | 6 | |
| Synthetic Image Detection | DALL-E 3 | AP0.756 | 6 | |
| Synthetic Image Detection | DALL-E 2 | AP86.7 | 6 | |
| Deepfake Detection | DALL-E mini | AP99.33 | 4 | |
| Deepfake Detection | DALL-E3 | Accuracy82 | 3 | |
| Text-to-Image Jailbreaking | DALL·E 3 with ChatGPT 5 | ASR48 | 3 | |
| Image Adherence | DALL-E 3 Prompt Set | Adherence Score (%)84.4 | 2 | |
| Image Quality | DALL-E 3 Prompt Set | Score90 | 2 |
Showing 24 of 24 rows