List your company

Text Classification on AGNews (Accuracy and p-value)

93.6AccuracyEspeW
73.00878.35483.789.046Apr 13, 2026Apr 19, 2026Apr 25, 2026May 1, 2026May 7, 2026May 13, 2026May 20, 2026

Evaluation results

MethodLinks
EspeWAttack=Dimension-shift Attack2026.0493.610
WARDENAttack=No Attack2026.0493.4810
GeoMarkAttack=Dimension-shift Attack2026.0493.4710
GeoMarkAttack=No Attack2026.0493.4510
OriginalAttack=None2026.0493.430.99
WARDENAttack=Dimension-shift Attack2026.0493.4310
EspeWAttack=No Attack2026.0493.4210
WARDENAttack=Dimension-reduction Attack2026.0493.3610
WETAttack=Paraphrasing Attack (NLLB)2026.0493.3210
WETAttack=Dimension-shift Attack2026.0493.270.92
GeoMarkAttack=Dimension-reduction Attack2026.0493.2410
WETAttack=No Attack2026.0493.2110
EspeWAttack=Dimension-reduction Attack2026.0493.210
WARDENAttack=CSE Attack2026.0492.9910
GeoMarkAttack=CSE Attack2026.0492.9710
EspeWAttack=CSE Attack2026.0492.9310
EspeWAttack=Paraphrasing Attack (GPT-4o-mini)2026.0492.8810
GeoMarkAttack=Paraphrasing Attack (GPT-4o-mini)2026.0492.8110
WARDENAttack=Paraphrasing Attack (GPT-4o-mini)2026.0492.7710
WETAttack=Paraphrasing Attack (GPT-4o-mini)2026.0492.6610
WARDENAttack=Paraphrasing Attack (NLLB)2026.0492.5410
GeoMarkAttack=Paraphrasing Attack (NLLB)2026.0492.510
EspeWAttack=Paraphrasing Attack (NLLB)2026.0492.4910
WETAttack=CSE Attack2026.0490.5310
ICLModel=LLaMA-2-13B2026.0588.04—
ICLModel=LLaMA-3.1-8B2026.0587.16—
LTVModel=LLaMA-2-13B2026.0586.68—
State VectorModel=LLaMA-2-13B2026.0584.64—
LTVModel=LLaMA-3.1-8B2026.0582.84—
Task VectorModel=LLaMA-3.1-8B2026.0581.36—
State VectorModel=LLaMA-3.1-8B2026.0580.28—
Task VectorModel=LLaMA-2-13B2026.0579.84—
I2CLModel=LLaMA-2-13B2026.0578.88—
I2CLModel=LLaMA-3.1-8B2026.0576.76—
Function VectorModel=LLaMA-3.1-8B2026.0576.16—
Zero-shotModel=LLaMA-3.1-8B2026.0575—
Function VectorModel=LLaMA-2-13B2026.0574—
Zero-shotModel=LLaMA-2-13B2026.0573.8—