List your company

Hallucination Detection on SQuAD v2

0.798ROC-AUCSinkProbe
0.630560.674030.71750.76097Apr 12, 2026

Evaluation results

MethodLinks
SinkProbeLLM=Llama3.1-8B2026.040.798
SinkProbeLLM=Mistral-Nemo2026.040.79
MTopDivLLM=Llama3.1-8B2026.040.787
LapEigvalLLM=Llama3.1-8B2026.040.785
LookbackLensLLM=Mistral-Nemo2026.040.783
MTopDivLLM=Phi3.52026.040.78
LapEigvalLLM=Mistral-Nemo2026.040.777
LookbackLensLLM=Llama3.1-8B2026.040.776
MTopDivLLM=Mistral-Nemo2026.040.77
SinkProbeLLM=Llama3.2-3B2026.040.769
LapEigvalLLM=Phi3.52026.040.767
LookbackLensLLM=Phi3.52026.040.767
LookbackLensLLM=Llama3.2-3B2026.040.765
MTopDivLLM=Llama3.2-3B2026.040.764
SinkProbeLLM=Phi3.52026.040.764
AttnEigvalLLM=Llama3.1-8B2026.040.761
AttnLogDetLLM=Phi3.52026.040.76
LapEigvalLLM=Llama3.2-3B2026.040.753
AttnLogDetLLM=Llama3.1-8B2026.040.752
AttnEigvalLLM=Phi3.52026.040.75
AttnLogDetLLM=Mistral-Nemo2026.040.749
AttnEigvalLLM=Mistral-Nemo2026.040.745
AttnLogDetLLM=Llama3.2-3B2026.040.739
AttnEigvalLLM=Llama3.2-3B2026.040.738
AttnScoreLLM=Phi3.52026.040.678
AttnScoreLLM=Llama3.1-8B2026.040.656
AttnScoreLLM=Mistral-Nemo2026.040.641
AttnScoreLLM=Llama3.2-3B2026.040.637