Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Watermark Detectability on C4 RealNewsLike Dipper-2 (test)
Loading...
85.9
AUC
MarkTune
53.868
62.184
70.5
78.816
Dec 3, 2025
AUC
TPR
Updated 1mo ago
Evaluation Results
Method
Method
Links
AUC
TPR
MarkTune
Backbone=Llama2-7B, Te...
2025.12
85.9
27
MarkTune
Backbone=Qwen3-4B, Tem...
2025.12
85.1
25.4
GaussMark
Backbone=Qwen3-4B, Tem...
2025.12
79.5
13.8
Gumbel-max
Backbone=Qwen3-4B, Tem...
2025.12
72.1
14.6
Gumbel-max
Backbone=Llama2-7B, Te...
2025.12
71.1
14.8
SynthID
Backbone=Llama2-7B, Te...
2025.12
71
8.4
SynthID
Backbone=Qwen3-4B, Tem...
2025.12
70
7.4
GaussMark
Backbone=Llama2-7B, Te...
2025.12
68.7
6
KGW-2
Backbone=Qwen3-4B, Tem...
2025.12
61.6
3.8
KGW-2
Backbone=Llama2-7B, Te...
2025.12
61.4
3.2
KGW-D
Backbone=Llama2-7B, Te...
2025.12
59
2.6
KGW-D
Backbone=Qwen3-4B, Tem...
2025.12
58.9
3
KGW-1
Backbone=Qwen3-4B, Tem...
2025.12
57.3
2.6
KGW-1
Backbone=Llama2-7B, Te...
2025.12
55.1
1.4
Feedback
Search any
task
Search any
task
