List your company

Question Answering on TruthfulQA o=1 (Exact split)

90AccuracyW/O Decontamination
26.24842.79959.3575.901Jan 27, 2026

Evaluation results

MethodLinks
W/O DecontaminationModel=LLaMA-3, Split=fcon, Access=-2026.01900.426
W/O DecontaminationModel=Mistral, Split=fcon, Access=-2026.01890.603
ITDModel=Mistral, Split=fcon, Access=Black-box2026.01890.603
Short CircuitModel=Mistral, Split=fcon, Access=White-box2026.01890.603
TEDModel=LLaMA-3, Split=fcon, Access=Black-box2026.0188.60.412
ITDModel=LLaMA-3, Split=fcon, Access=Black-box2026.0187.60.402
W/O DecontaminationModel=Qwen 2.5, Split=fcon, Access=-2026.0186.60.311
ITDModel=Qwen 2.5, Split=fcon, Access=Black-box2026.0185.60.301
Shortcut NeuronModel=LLaMA-3, Split=fcon, Access=White-box2026.0180.40.33
TEDModel=Qwen 2.5, Split=fcon, Access=Black-box2026.0178.90.234
TEDModel=Mistral, Split=fcon, Access=Black-box2026.0178.50.498
Shortcut NeuronModel=Mistral, Split=fcon, Access=White-box2026.01630.343
DeconIEPModel=Qwen 2.5, Split=fcon, Access=White-box2026.0161.80.063
Short CircuitModel=Qwen 2.5, Split=fcon, Access=White-box2026.0160.20.047
Short CircuitModel=LLaMA-3, Split=fcon, Access=White-box2026.0158.90.115
baselineModel=Qwen 2.5, Split=fclean, Access=-2026.0155.5—
DeconIEPModel=LLaMA-3, Split=fcon, Access=White-box2026.0151.70.043
baselineModel=LLaMA-3, Split=fclean, Access=-2026.0147.4—
DeconIEPModel=Mistral, Split=fcon, Access=White-box2026.0136.60.079
Shortcut NeuronModel=Qwen 2.5, Split=fcon, Access=White-box2026.0130.60.249
baselineModel=Mistral, Split=fclean, Access=-2026.0128.7—