Watermark Detection on Ten-exam benchmark 1.0 (test)
92.8Detection ScoreIS-v2
Evaluation results
| Method | Links | |
|---|---|---|
| IS-v2Target Model=GPT2026.01 | 92.8 | |
| IS-v2Target Model=Claude2026.01 | 92.1 | |
| IS-v2Target Model=Grok2026.01 | 91.6 | |
| IS-v2Target Model=Gemini2026.01 | 91 | |
| IS-v1Target Model=GPT2026.01 | 90.7 | |
| IS-v1Target Model=Claude2026.01 | 90.3 | |
| IS-v1Target Model=Grok2026.01 | 89.9 | |
| IS-v1Target Model=Gemini2026.01 | 89.5 | |
| TRAPDOCTarget Model=GPT2026.01 | 87.9 | |
| code-glyphTarget Model=GPT2026.01 | 85.9 | |
| TRAPDOCTarget Model=Claude2026.01 | 85.8 | |
| TRAPDOCTarget Model=Grok2026.01 | 84.6 | |
| code-glyphTarget Model=Claude2026.01 | 84.2 | |
| code-glyphTarget Model=Grok2026.01 | 82.7 | |
| code-glyphTarget Model=Gemini2026.01 | 81.4 | |
| TRAPDOCTarget Model=Gemini2026.01 | 43.4 | |
| ICWTarget Model=GPT2026.01 | 6.8 | |
| ICWTarget Model=Claude2026.01 | 5.3 | |
| ICWTarget Model=Grok2026.01 | 4.6 | |
| ICWTarget Model=Gemini2026.01 | 3.2 |