List your company

Question Answering on TruthfulQA Exact split, o=3

98.6AccuracyW/O Decontamination
25.90444.77763.6582.523Jan 27, 2026

Evaluation results

MethodLinks
W/O DecontaminationModel=LLaMA-3, Split=f_con2026.0198.651.2
W/O DecontaminationModel=Mistral, Split=f_con2026.0197.668.9
ITDModel=Mistral, Split=f_con, Access=Black-box2026.0197.668.9
W/O DecontaminationModel=Qwen2.5, Split=f_con2026.0197.642.1
ITDModel=Qwen2.5, Split=f_con, Access=Black-box2026.0197.642.1
TEDModel=LLaMA-3, Split=f_con, Access=Black-box2026.0196.649.3
ITDModel=LLaMA-3, Split=f_con, Access=Black-box2026.0194.346.9
Short CircuitModel=Mistral, Split=f_con, Access=White-box2026.0193.364.6
TEDModel=Qwen2.5, Split=f_con, Access=Black-box2026.018832.5
Shortcut NeuronModel=LLaMA-3, Split=f_con, Access=White-box2026.0181.834.5
Shortcut NeuronModel=Qwen2.5, Split=f_con, Access=White-box2026.0168.412.9
Shortcut NeuronModel=Mistral, Split=f_con, Access=White-box2026.0166.838.1
Short CircuitModel=Qwen2.5, Split=f_con, Access=White-box2026.01659.5
DeconIEPModel=Qwen2.5, Split=f_con, Access=White-box2026.01626.5
Short CircuitModel=LLaMA-3, Split=f_con, Access=White-box2026.0159.312
baselineModel=Qwen2.5, Split=f_clean2026.0155.5—
DeconIEPModel=LLaMA-3, Split=f_con, Access=White-box2026.0153.66.2
TEDModel=Mistral, Split=f_con, Access=Black-box2026.0153.524.8
DeconIEPModel=Mistral, Split=f_con, Access=White-box2026.0153.324.6
baselineModel=LLaMA-3, Split=f_clean2026.0147.4—
baselineModel=Mistral, Split=f_clean2026.0128.7—