List your company

CoT Faithfulness Detection on Truthful-QA

78AccuracyCIE-SCORER
35.04846.19957.3568.501May 25, 2026

Evaluation results

MethodLinks
CIE-SCORERParadigm=Circuit-based2026.057871.5
CRVParadigm=Circuit-based2026.056866.7
ParaphrasingParadigm=Counterfactual-based2026.0567.849.1
BiGGenParadigm=LLM-as-judge2026.0561.167.3
Adding MistakesParadigm=Counterfactual-based2026.0551.160.7
Option ShufflingParadigm=Counterfactual-based2026.0551.159.3
Information GainParadigm=Logits-based2026.0547.840.5
PerplexityParadigm=Baselines2026.0544.440.5
RandomParadigm=Baselines2026.0543.342.7
Early AnsweringParadigm=Counterfactual-based2026.054052.6
Answer TracingParadigm=Logits-based2026.0538.950.5
Removing StepsParadigm=Counterfactual-based2026.0536.750.4