Text Classification on SST-2 (Accuracy and P-value)
93.62AccuracyGeoMark
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GeoMarkAttack=Dimension-shift Attack2026.04 | 93.62 | 10 | |
| OriginalAttack=None2026.04 | 93.54 | 0.38 | |
| WARDENAttack=No Attack2026.04 | 93.5 | 10 | |
| GeoMarkAttack=Paraphrasing Attack (NLLB)2026.04 | 93.46 | 0.002 | |
| EspeWAttack=Dimension-shift Attack2026.04 | 93.39 | 10 | |
| EspeWAttack=No Attack2026.04 | 93.35 | 10 | |
| GeoMarkAttack=Dimension-reduction Attack2026.04 | 93.35 | 10 | |
| WARDENAttack=Paraphrasing Attack (NLLB)2026.04 | 93.31 | 10 | |
| WARDENAttack=Dimension-shift Attack2026.04 | 93.27 | 10 | |
| WETAttack=No Attack2026.04 | 93.27 | 10 | |
| WETAttack=Paraphrasing Attack (NLLB)2026.04 | 93.27 | 10 | |
| WETAttack=Dimension-shift Attack2026.04 | 93.23 | 0.8 | |
| GeoMarkAttack=No Attack2026.04 | 93.23 | 10 | |
| EspeWAttack=Paraphrasing Attack (NLLB)2026.04 | 93.16 | 0.11 | |
| EspeWAttack=Dimension-reduction Attack2026.04 | 93.16 | 10 | |
| WARDENAttack=Dimension-reduction Attack2026.04 | 93 | 10 | |
| WETAttack=Paraphrasing Attack (GPT-4o-mini)2026.04 | 92.78 | 10 | |
| GeoMarkAttack=Paraphrasing Attack (GPT-4o-mini)2026.04 | 92.43 | 10 | |
| EspeWAttack=Paraphrasing Attack (GPT-4o-mini)2026.04 | 92.39 | 0.11 | |
| WARDENAttack=Paraphrasing Attack (GPT-4o-mini)2026.04 | 92.16 | 0.15 | |
| WARDENAttack=CSE Attack2026.04 | 89.03 | 10 | |
| GeoMarkAttack=CSE Attack2026.04 | 88.23 | 10 | |
| WETAttack=CSE Attack2026.04 | 87.19 | 10 | |
| EspeWAttack=CSE Attack2026.04 | 86.85 | 10 |