Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Deepfake

Benchmarks

Task NameDataset NameSOTA ResultTrend
Image ClassificationDeepFake JSD = 0.545 (GS)
TP643
20
Image ClassificationDeepFake JSD = 1e-4 (RS)
True Positives (TP)793
20
Tamper LocalizationDeepFake
IoU39
12
AI-generated image detectionDeepfake dataset
Accuracy89.71
11
AI-generated text detectionDeepfake In-distribution, Cross-domains & Cross-models
Average Recall96.15
10
Fake image detectionDeepfake
AA99.06
9
Fake Image DetectionDeepfake (test)
AP92.82
6
AI-generated text detectionDeepfake Cross-domains & Cross-models (test)
Avg Recall0.9615
6
AI-generated text detectionDeepfake Out-of-distribution Unseen Domains
Avg Recall89.63
6
AI-generated text detectionDeepfake Out-of-distribution Unseen Models
Avg Recall0.9303
6
AI-generated text detectionDeepfake Domain-specific & Model-specific (In-distribution)
Average Recall99.77
5
AI-generated text detectionDeepfake Cross-domains Model-specific (In-distribution)
Avg Recall96.73
5
Showing 12 of 12 rows