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

Watermarking on Natural Questions (NQ) (test)

100AUROC

KGW

Updated 5d ago

Evaluation Results

Method	Links
KGW 2026.05		100	100	100
SynthID 2026.05		100	100	100
k-SemStamp 2026.05		100	96.3	98.4
SimMark 2026.05		100	99.2	99.4
AliMark 2026.05		100	100	100
AliMark 2026.05		99.9	93.8	97.2
AliMark 2026.05		99.9	93.7	96.9
SimMark 2026.05		99.8	93.6	95.8
SemStamp 2026.05		99.7	94.1	97
SIR 2026.05		99.5	0	91
SimMark 2026.05		98.7	90.6	93.2
PMark 2026.05		98.6	92.2	94.6
EXP 2026.05		98	96.4	97
SemStamp 2026.05		97.6	58	79.4
PMark 2026.05		97.1	47	81.8
SemStamp 2026.05		97	41.4	70.8
k-SemStamp 2026.05		97	45.3	54.5
SIR 2026.05		96.6	0	33.6
k-SemStamp 2026.05		96	37.3	47.1
PMark 2026.05		95.8	48	79
SIR 2026.05		95.7	0	33.8
AliMark 2026.05		95.1	54.7	59.8
SimMark 2026.05		93.6	40.6	53.6
AliMark 2026.05		91.7	49.5	55.2
SIR 2026.05		90.7	0	4
SimMark 2026.05		90.1	42.2	50.3
SynthID 2026.05		86.9	8.2	18.2
SynthID 2026.05		86.4	6.2	14
SIR 2026.05		84.5	0	5.4
SemStamp 2026.05		84.5	10.3	24.7
SemStamp 2026.05		79.5	5.2	17.6
k-SemStamp 2026.05		79.5	2.6	5.8
PMark 2026.05		77.1	5.6	17.6
k-SemStamp 2026.05		76.4	1.8	3.8
PMark 2026.05		75.9	5.4	16.2
KGW 2026.05		61.6	0.6	1.8
KGW 2026.05		61.2	0.6	2
EXP 2026.05		59.2	2.4	4.6
KGW 2026.05		59.1	2.2	3.2
SynthID 2026.05		57.9	0	0
KGW 2026.05		57.6	0.4	1.4
EXP 2026.05		57.2	1.2	3
EXP 2026.05		55.6	0.8	2.6
EXP 2026.05		55.6	0.2	0.8
SynthID 2026.05		55	0	0