List your company

Sentiment Analysis on SST2 (Adversarial Attack Evaluation)

2.1Attack Success Rate (ASR)LwF
-1.81624.61751.0577.483Nov 27, 2025Nov 29, 2025Dec 1, 2025Dec 3, 2025Dec 5, 2025Dec 7, 2025Dec 10, 2025

Evaluation results

MethodLinks
LwFPoisoned fine-tuning=p=0.1, Backbone=Safety-aligned LLaMA2-7B2025.122.194.5
DERPoisoned fine-tuning=p=0.1, Backbone=Safety-aligned LLaMA2-7B2025.125.994.3
SafeInstrPoisoned fine-tuning=p=0.1, Backbone=Safety-aligned LLaMA2-7B2025.126.494.7
LisaPoisoned fine-tuning=p=0.1, Backbone=Safety-aligned LLaMA2-7B2025.126.695.1
A-GEMPoisoned fine-tuning=p=0.1, Backbone=Safety-aligned LLaMA2-7B2025.121594.6
SafeLoRAPoisoned fine-tuning=p=0.1, Backbone=Safety-aligned LLaMA2-7B2025.1272.494.7
SFTPoisoned fine-tuning=p=0.1, Backbone=Safety-aligned LLaMA2-7B2025.1278.394.5
CacheTrapModel=Mistral-7B, Target Class=12025.1195.3—
CacheTrapModel=Qwen-2.5-3B, Target Class=12025.1199.89—
CacheTrapModel=DeepSeek-7B, Target Class=12025.1199.89—
CacheTrapModel=LLaMA-2-7B, Target Class=02025.1110096.44
CacheTrapModel=LLaMA-2-7B, Target Class=12025.11100—
CacheTrapModel=LLaMA-3.1-8B, Target Class=02025.1110096.67
CacheTrapModel=LLaMA-3.1-8B, Target Class=12025.11100—
CacheTrapModel=Mistral-7B, Target Class=02025.1110096.22
CacheTrapModel=Qwen-2.5-3B, Target Class=02025.1110095.87
CacheTrapModel=DeepSeek-7B, Target Class=02025.1110095.3