List your company

Internal-trigger targeted bit-flip attack on WebShop (test)

0.95CDAFlip-Agent
0.2220.4110.60.789Mar 7, 2026

Evaluation results

MethodLinks
Flip-AgentModel=AgentLM-7B2026.030.950.944
Flip-AgentModel=DeepSeek-R1-Distill-Llama-8B2026.030.950.222
Flip-SModel=AgentLM-7B2026.030.90.722
Flip-AgentModel=AgentLM-13B2026.030.90.444
Flip-SModel=DeepSeek-R1-Distill-Llama-8B2026.030.90.111
TrojViTModel=AgentLM-7B2026.030.850.509
Flip-SModel=AgentLM-13B2026.030.850.175
TrojViTModel=DeepSeek-R1-Distill-Llama-8B2026.030.830.079
TrojViTModel=Llama-3.2-3B-Instruct2026.030.80.273
Flip-AgentModel=Llama-3.2-3B-Instruct2026.030.80.5
TBTModel=AgentLM-7B2026.030.80.167
Flip-AgentModel=Qwen3-4B-Instruct-25072026.030.80.389
Flip-SModel=Llama-3.2-3B-Instruct2026.030.750.204
Flip-AgentModel=Llama-3.1-8B-Instruct2026.030.750.611
TBTModel=Qwen3-4B-Instruct-25072026.030.750.111
TrojViTModel=AgentLM-13B2026.030.70.116
TrojViTModel=Qwen3-4B-Instruct-25072026.030.70.13
Flip-SModel=Qwen3-4B-Instruct-25072026.030.70.222
TBTModel=Llama-3.2-3B-Instruct2026.030.650.111
TrojViTModel=Llama-3.1-8B-Instruct2026.030.60.224
TBTModel=DeepSeek-R1-Distill-Llama-8B2026.030.50.056
Flip-SModel=Llama-3.1-8B-Instruct2026.030.450.389
TBTModel=AgentLM-13B2026.030.30
TBTModel=Llama-3.1-8B-Instruct2026.030.250.073