Fraud on Product
100ASRTIP
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| TIPDefense=No Defense2026.03 | 100 | 100 | |
| TIPDefense=Instruction Prevention2026.03 | 100 | 20 | |
| TIPDefense=Sandwich Prevention2026.03 | 100 | 260 | |
| TIPDefense=Perplexity Filtering2026.03 | 100 | 40 | |
| TAPDefense=Instruction Prevention2026.03 | 99 | 660 | |
| TAPDefense=No Defense2026.03 | 91 | 2,580 | |
| TIPDefense=Finetuned Detector2026.03 | 90 | 140 | |
| TAPDefense=Sandwich Prevention2026.03 | 78.6 | 480 | |
| TAPDefense=Perplexity Filtering2026.03 | 52.5 | 2,480 | |
| TAPDefense=Finetuned Detector2026.03 | 46.5 | 2,580 | |
| FixedDefense=Perplexity Filtering2026.03 | 28 | — | |
| FixedDefense=Finetuned Detector2026.03 | 17 | — | |
| FixedDefense=No Defense2026.03 | 16 | — | |
| FixedDefense=Instruction Prevention2026.03 | 10 | — | |
| FixedDefense=Sandwich Prevention2026.03 | 5 | — |