DALL-E
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
DALL·E (test)
29.4TPR @ 5% FPR
54
DALL-E 2 (test)
0.826Average Similarity
28
DALL-E
72.3Accuracy
26
DALL-E 3
98.1AP
15
DALL-E 2 (full)
77.3Acc@EER
12
DALL·E
98.81AA
9
DALL-E 2 (Academic)
0.727AUCROC
9
DALL-E 3 (proprietary)
0.759AUCROC
9
DALL-E 2 (proprietary)
85.4AUCROC
9
DALL-E
98.8Accuracy
9
DALL·E Universal nudity jailbreaking prompts 3
35.58Transfer Success Rate
7
DALL-E 3
18.28PSNR
7
DALL-E 3
100Robustness: Brightness
7
DALL·E 3 commercial (test)
0.95BR
7
DALL·E 3
12.98TASR (%)
6
DALL-E 3 (full)
69.8Acc@EER
6
DALL-E Anime 2
62AP
6
DALL-E 3
0.756AP
6
DALL-E 2
86.7AP
6
DALL-E mini
99.33AP
4
DALL-E3
82Accuracy
3
DALL·E 3 with ChatGPT 5
48ASR
3
DALL-E 3 Prompt Set
84.4Adherence Score (%)
2
DALL-E 3 Prompt Set
90Score
2