| Dataset Name | SOTA Method | Metric | Trend | ||
|---|---|---|---|---|---|
| Stable Diffusion-Prompts (SDP) 350 watermarked images | ZoDiac | TPR@1%FPR100 | 108 | 3d ago | |
| Vicuna-7b 16k 50 samples v1.5 | RP4 | AUROC (Overall)0.986 | 94 | 3d ago | |
| ImageNet 2014 (val) | Detection Rate (Level 1)100 | 66 | 3d ago | ||
| Llama-2-7b-chat-hf 10 samples UMD watermarking (test) | RP1 | AUROC (t=0)1 | 64 | 3d ago | |
| finance_qa | MC2MARK | Accuracy100 | 48 | 3d ago | |
| longform_qa | MC2MARK | Accuracy100 | 48 | 3d ago | |
| dolly_cw | MC2MARK | Accuracy100 | 48 | 3d ago | |
| fake_news | MC2MARK | Accuracy100 | 48 | 3d ago | |
| mmw story | MC2MARK | Accuracy100 | 48 | 3d ago | |
| book_report | MC2MARK | Accuracy100 | 48 | 3d ago | |
| ImageNet | GS. | Robustness - Scaling99.51 | 33 | 3d ago | |
| GSM8K | He et al. | True Detection Rate (TD)100 | 30 | 3d ago | |
| c4 subset | MC2MARK | Accuracy100 | 24 | 3d ago | |
| C4 subset | MC2MARK | Accuracy100 | 24 | 3d ago | |
| Ten-exam benchmark 1.0 (test) | IS-v2 | Detection Score92.8 | 20 | 3d ago | |
| VGGFace | AccLoss6.93 | 14 | 3d ago | ||
| CIFAR100 | ComMark | AccLoss5.23 | 14 | 3d ago | |
| CIFAR10 | ComMark | AccLoss8.13 | 14 | 3d ago | |
| GTSRB | ComMark | AccLoss12.29 | 14 | 3d ago | |
| Llama-3-8B-Instruct 150 tokens (generations) | KGW | Mean P9 | 13 | 3d ago | |
| Llama-3 8B Instruct 30 tokens (generations) | KGW | Mean Precision23 | 13 | 3d ago | |
| C4 250 tokens | ENS-MCMark | TPR @ FPR 0.1%96.7 | 12 | 3d ago | |
| C4 150 tokens | ENS-MCMark | TPR @ FPR 0.1%0.822 | 12 | 3d ago | |
| SimpleQA | ACTHOOK | Delta_q0.81 | 10 | 3d ago | |
| MATH | ACTHOOK | AUC (Unspecified Config)99.9 | 10 | 3d ago |