Watermark Detection on MBPP
97.1TPR @ 1% FPREWD + SSG
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| EWD + SSGModel=LLaMA-3-8B2026.04 | 97.1 | — | — | — | 98.1 | 98.9 | 97.1 | |
| SWEET + SSGModel=LLaMA-3-8B2026.04 | 93 | — | — | — | 96 | 96.8 | 96.1 | |
| EWDModel=LLaMA-3-8B2026.04 | 87.7 | — | — | — | 91.3 | 94.2 | 94.7 | |
| KGW + SSGModel=LLaMA-3-8B2026.04 | 85.5 | — | — | — | 89.9 | 89.9 | 92.4 | |
| EWD + SSGModel=Qwen2.5-coder-7B2026.04 | 74.3 | — | — | — | 84.9 | 84.1 | 89.1 | |
| SWEETModel=LLaMA-3-8B2026.04 | 74.1 | — | — | — | 78.5 | 81 | 87.2 | |
| SWEET + SSGModel=Qwen2.5-coder-7B2026.04 | 62.2 | — | — | — | 83.5 | 76.2 | 84.2 | |
| KGW + SSGModel=Qwen2.5-coder-7B2026.04 | 58.7 | — | — | — | 73.6 | 71.7 | 81.3 | |
| EWDModel=Qwen2.5-coder-7B2026.04 | 57.4 | — | — | — | 75.7 | 74.6 | 83.2 | |
| SWEETModel=Qwen2.5-coder-7B2026.04 | 53.7 | — | — | — | 69.5 | 72.8 | 82 | |
| KGWModel=Qwen2.5-coder-7B2026.04 | 37.8 | — | — | — | 54.6 | 64 | 75.9 | |
| KGWModel=LLaMA-3-8B2026.04 | 21.1 | — | — | — | 27.6 | 28.3 | 42.5 | |
| UWModel=LLaMA-3-8B2026.04 | 13.6 | — | — | — | 18.3 | 29.7 | 10.5 | |
| UWModel=Qwen2.5-coder-7B2026.04 | 1.9 | — | — | — | 3.6 | 5.8 | 10.5 | |
| EWDScenario=Low-entropy, FPR threshold=1%2024.03 | — | 56.7 | 72 | — | — | — | — | |
| EWDScenario=Low-entropy, FPR threshold=5%2024.03 | — | 74.4 | 83 | — | — | — | — | |
| EWDScenario=Low-entropy, FPR threshold=Best2024.03 | — | — | 87.8 | — | — | — | — | |
| EXP-editWatermarking Status=Watermarked2026.01 | — | — | — | 59.8 | — | — | — | |
| KGWScenario=Low-entropy, FPR threshold=1%2024.03 | — | 11.5 | 20.5 | — | — | — | — | |
| KGWScenario=Low-entropy, FPR threshold=5%2024.03 | — | 34.8 | 49.8 | — | — | — | — | |
| KGWScenario=Low-entropy, FPR threshold=Best2024.03 | — | — | 74.4 | — | — | — | — | |
| SWaRLWatermarking Status=Watermarked2026.01 | — | — | — | 90.8 | — | — | — | |
| SWEETScenario=Low-entropy, FPR threshold=1%2024.03 | — | 40.9 | 57.7 | — | — | — | — | |
| SWEETScenario=Low-entropy, FPR threshold=5%2024.03 | — | 65 | 76.5 | — | — | — | — | |
| SWEETScenario=Low-entropy, FPR threshold=Best2024.03 | — | — | 86.5 | — | — | — | — | |
| SWEETWatermarking Status=Watermarked2026.01 | — | — | — | 89.8 | — | — | — | |
| WLLMWatermarking Status=Watermarked2026.01 | — | — | — | 89 | — | — | — |