List your company

Language Understanding on MMLU (o=1 Exact Split)

77.6AccuracyITD
24.24838.09951.9565.801Jan 27, 2026

Evaluation results

MethodLinks
ITDModel=Qwen 2.5, Split=fcon, Access=Black-box2026.0177.60.106
W/O DecontaminationModel=Qwen 2.5, Split=fcon, Access=-2026.0177.50.105
TEDModel=Qwen 2.5, Split=fcon, Access=Black-box2026.0174.20.072
ITDModel=Qwen 2.5, Split=fcon, Access=Black-box2026.0174.20.069
W/O DecontaminationModel=Qwen 2.5, Split=fcon, Access=-2026.0173.50.062
W/O DecontaminationModel=LLaMA-3, Split=fcon, Access=-2026.0173.40.159
DeconIEPModel=Qwen 2.5, Split=fcon, Access=White-box2026.01730.057
TEDModel=LLaMA-3, Split=fcon, Access=Black-box2026.0172.70.151
ITDModel=LLaMA-3, Split=fcon, Access=Black-box2026.0172.10.145
Short CircuitModel=LLaMA-3, Split=fcon, Access=White-box2026.01710.16
DeconIEPModel=Qwen 2.5, Split=fcon, Access=White-box2026.0170.50.035
Short CircuitModel=LLaMA-3, Split=fcon, Access=White-box2026.0169.50.12
baselineModel=Qwen 2.5, Split=fclean, Access=-2026.0167.3—
baselineModel=Qwen 2.5, Split=fclean, Access=-2026.0167—
W/O DecontaminationModel=LLaMA-3, Split=fcon, Access=-2026.0166.30.114
Short CircuitModel=Mistral, Split=fcon, Access=White-box2026.0163.10.18
ITDModel=LLaMA-3, Split=fcon, Access=Black-box2026.01620.07
Short CircuitModel=Qwen 2.5, Split=fcon, Access=White-box2026.0161.90.051
TEDModel=Qwen 2.5, Split=fcon, Access=Black-box2026.0161.90.054
W/O DecontaminationModel=Mistral, Split=fcon, Access=-2026.0161.70.166
DeconIEPModel=LLaMA-3, Split=fcon, Access=White-box2026.0161.60.04
ITDModel=Mistral, Split=fcon, Access=Black-box2026.01610.159
Short CircuitModel=Qwen 2.5, Split=fcon, Access=White-box2026.01610.063
Short CircuitModel=Mistral, Split=fcon, Access=White-box2026.0159.40.188
baselineModel=LLaMA-3, Split=fclean, Access=-2026.0157.6—
TEDModel=LLaMA-3, Split=fcon, Access=Black-box2026.0157.60.027
baselineModel=LLaMA-3, Split=fclean, Access=-2026.0154.9—
Shortcut NeuronModel=LLaMA-3, Split=fcon, Access=White-box2026.0154.70.029
DeconIEPModel=LLaMA-3, Split=fcon, Access=White-box2026.0151.50.034
ITDModel=Mistral, Split=fcon, Access=Black-box2026.0149.90.093
W/O DecontaminationModel=Mistral, Split=fcon, Access=-2026.0149.20.086
Shortcut NeuronModel=LLaMA-3, Split=fcon, Access=White-box2026.0148.40.065
DeconIEPModel=Mistral, Split=fcon, Access=White-box2026.0147.80.027
DeconIEPModel=Mistral, Split=fcon, Access=White-box2026.01470.064
baselineModel=Mistral, Split=fclean, Access=-2026.0145.1—
baselineModel=Mistral, Split=fclean, Access=-2026.0140.6—
TEDModel=Mistral, Split=fcon, Access=Black-box2026.0140.40.047
Shortcut NeuronModel=Qwen 2.5, Split=fcon, Access=White-box2026.0139.50.275
Shortcut NeuronModel=Qwen 2.5, Split=fcon, Access=White-box2026.0138.90.284
TEDModel=Mistral, Split=fcon, Access=Black-box2026.0136.30.043
Shortcut NeuronModel=Mistral, Split=fcon, Access=White-box2026.01300.151
Shortcut NeuronModel=Mistral, Split=fcon, Access=White-box2026.0126.30.143