List your company

Text Classification on Amazon (ACC, ASR)

89.71AccuracyBadTemplate
-3.03221.0452545.122569.19975Feb 5, 2026

Evaluation results

MethodLinks
BadTemplateModel=Deepseek-v3.2, Attack Level=word-level2026.0289.7197.66
GPT-4o-miniProtocol=ICL-1Shot, Attack Level=Sentence-level2026.0289.1610.83
Deepseek-v3.2Protocol=Clean, Attack Level=Sentence-level2026.0289.1614.25
BadTemplateBackbone=Deepseek-v3.2, Attack Level=Sentence-level2026.0289.1697.5
CleanModel=Deepseek-v3.2, Attack Level=word-level2026.0289.1614.25
Deepseek-v3.2Protocol=ICL-3Shot, Attack Level=Sentence-level2026.0289.0811
GPT-4o-miniProtocol=ICL-3Shot, Attack Level=Sentence-level2026.028911
Deepseek-v3.2Protocol=ICL-1Shot, Attack Level=Sentence-level2026.0288.9110.41
BadTemplateBackbone=Gemini-2.5, Attack Level=Sentence-level2026.0288.75100
GPT-4o-miniProtocol=ICL-2Shot, Attack Level=Sentence-level2026.0288.7510.25
Deepseek-v3.2Protocol=ICL-2Shot, Attack Level=Sentence-level2026.0288.6610.08
GPT-4o-miniProtocol=Clean, Attack Level=Sentence-level2026.0288.516.67
BadTemplateBackbone=GPT-4o-mini, Attack Level=Sentence-level2026.0288.591.16
CleanModel=GPT-4o-mini, Attack Level=word-level2026.0288.516.67
BadTemplateModel=Gemini-2.5, Attack Level=word-level2026.0288.4199.83
Gemini-2.5Protocol=ICL-1Shot, Attack Level=Sentence-level2026.0288.3314.25
Gemini-2.5Protocol=Clean, Attack Level=Sentence-level2026.0288.2516.67
CleanModel=Gemini-2.5, Attack Level=word-level2026.0288.2516.67
Gemini-2.5Protocol=ICL-2Shot, Attack Level=Sentence-level2026.0287.8313.91
BaselineModel=Llama-70B2026.0287.7516.67
BadTemplateModel=GPT-4o-mini, Attack Level=word-level2026.0287.7592.5
BadTemplateModel=Llama-70B2026.0287.2535.58
ICL-1shotModel=Llama-70B2026.0286.917.83
ICL-2shotModel=Llama-70B2026.0286.837.66
ICL-3shotModel=Llama-70B2026.0286.757.41
BaselineModel=Yi-34B2026.0286.3316.67
BadTemplateModel=Yi-34B2026.0285.6647.25
ICL-2shotModel=Llama-8B2026.028514.75
ICL-1shotModel=Yi-34B2026.0284.418.33
ICL-2shotModel=Yi-34B2026.0283.757.25
ICL-1shotModel=Llama-8B2026.028314.58
ICL-3shotModel=Yi-34B2026.0282.9111
BadTemplateModel=Mistral-Small2026.0282.6616.58
Gemini-2.5Protocol=ICL-3Shot, Attack Level=Sentence-level2026.0281.7531.91
BaselineModel=Mistral-Small2026.0281.0816.67
ICL-3shotModel=Llama-8B2026.028119.41
BadTemplateModel=Llama-8B2026.0280.0846.33
BaselineModel=Llama-8B2026.0279.516.67
ICL-2shotModel=Mistral-Small2026.0275.4115
BaselineModel=Phi-3.52026.0275.1616.67
BadTemplateModel=Phi-3.52026.0274.1630.66
ICL-1shotModel=Mistral-Small2026.0270.6612.66
BaselineModel=Deepseek-7B2026.0262.6616.67
ICL-1shotModel=Deepseek-7B2026.0261.7522.58
BadTemplateModel=Deepseek-7B2026.0261.7578.83
ICL-2shotModel=Deepseek-7B2026.0261.5817.16
ICL-3shotModel=Deepseek-7B2026.0259.0816.75
ICL-3shotModel=Mistral-Small2026.0258.1612.83
ICL-3shotModel=Phi-3.52026.0255.088.08
ICL-1shotModel=Phi-3.52026.02556.83
ICL-2shotModel=Phi-3.52026.0252.586.75
CleanBackbone=Llama-70B2026.020.87750.1667
BadTemplateBackbone=Llama-70B2026.020.86750.1325
ICL-1ShotBackbone=Llama-70B2026.020.86680.075
ICL-2ShotBackbone=Llama-70B2026.020.86660.0733
ICL-3ShotBackbone=Llama-70B2026.020.86580.0725
CleanBackbone=Yi-34B2026.020.86330.1667
BadTemplateBackbone=Yi-34B2026.020.85410.7291
ICL-1ShotBackbone=Yi-34B2026.020.85160.0816
ICL-2ShotBackbone=Yi-34B2026.020.84830.0883
ICL-3ShotBackbone=Yi-34B2026.020.83830.0966
ICL-1ShotBackbone=Llama-8B2026.020.82830.1258
ICL-2ShotBackbone=Llama-8B2026.020.81750.1325
BadTemplateBackbone=Llama-8B2026.020.81410.3208
BadTemplateBackbone=Mistral-Small2026.020.81160.0566
CleanBackbone=Mistral-Small2026.020.81080.1667
ICL-3ShotBackbone=Llama-8B2026.020.80160.1625
CleanBackbone=Llama-8B2026.020.7950.1667
ICL-2ShotBackbone=Mistral-Small2026.020.75830.1258
CleanBackbone=Phi-3.52026.020.75160.1667
ICL-1ShotBackbone=Mistral-Small2026.020.7350.1233
BadTemplateBackbone=Phi-3.52026.020.72910.2066
CleanBackbone=Deepseek-7B2026.020.62660.1667
BadTemplateBackbone=Deepseek-7B2026.020.62080.6958
ICL-2ShotBackbone=Deepseek-7B2026.020.620.1941
ICL-1ShotBackbone=Deepseek-7B2026.020.61410.2125
ICL-3ShotBackbone=Deepseek-7B2026.020.61080.1775
ICL-3ShotBackbone=Mistral-Small2026.020.58910.1158
ICL-3ShotBackbone=Phi-3.52026.020.57830.0558
ICL-1ShotBackbone=Phi-3.52026.020.55410.0458
ICL-2ShotBackbone=Phi-3.52026.020.5350.0333