List your company

Instruction Following on Alpaca poisoned (test)

99.93F1 ScoreSyns
43.780458.357772.93587.5123May 12, 2026

Evaluation results

MethodLinks
SynsLLM Architecture=Mistral 7B2026.0599.9399.8
ParaesthesiaLLM Architecture=Qwen2.5 7B Instruct2026.0599.8799.75
ParaesthesiaLLM Architecture=Mistral 7B2026.0599.8299.6
ParaesthesiaLLM Architecture=Llama 2 7B2026.0599.7799.55
SynsLLM Architecture=Qwen2.5 7B Instruct2026.0599.6899.35
SynsLLM Architecture=Vicuna 7B2026.0599.6799.2
ParaesthesiaLLM Architecture=Vicuna 7B2026.0599.6699.3
SleeperLLM Architecture=Mistral 7B2026.0599.6499.25
SleeperLLM Architecture=Llama 2 7B2026.0599.6399.15
VPILLM Architecture=Llama 2 7B2026.0599.5599.05
SynsLLM Architecture=Llama 2 7B2026.0599.5299.05
VPILLM Architecture=Vicuna 7B2026.0599.4198.75
VPILLM Architecture=Qwen2.5 7B Instruct2026.0599.4198.75
SleeperLLM Architecture=Qwen2.5 7B Instruct2026.0599.3898.7
SyndLLM Architecture=Llama 2 7B2026.0599.3798.75
SleeperLLM Architecture=Vicuna 7B2026.0599.3398.6
SyndLLM Architecture=Mistral 7B2026.0599.2998.55
VPILLM Architecture=Mistral 7B2026.0599.298.35
SyndLLM Architecture=Vicuna 7B2026.0599.0798.05
SyndLLM Architecture=Qwen2.5 7B Instruct2026.0598.3996.75
CBALLM Architecture=Qwen2.5 7B Instruct2026.0595.6491.65
CBALLM Architecture=Llama 2 7B2026.0594.7389.95
CBALLM Architecture=Mistral 7B2026.0594.4689.3
CBALLM Architecture=Vicuna 7B2026.0592.6285.9
BadNetsLLM Architecture=Llama 2 7B2026.0561.741
BadNetsLLM Architecture=Mistral 7B2026.0546.450
BadNetsLLM Architecture=Vicuna 7B2026.0546.140.05
BadNetsLLM Architecture=Qwen2.5 7B Instruct2026.0545.940.05