List your company

CoT faithfulness detection on AQuA

77Accuracy (CoT Faithfulness)CIE-SCORER
40.18449.74259.368.858May 25, 2026

Evaluation results

MethodLinks
CIE-SCORERParadigm=Circuit-based2026.057772.8
BiGGenParadigm=LLM-as-judge2026.0573.770.3
Removing StepsParadigm=Counterfactual-based2026.0572.446.2
Early AnsweringParadigm=Counterfactual-based2026.0572.453.3
Adding MistakesParadigm=Counterfactual-based2026.0571.666.7
ParaphrasingParadigm=Counterfactual-based2026.0568.442.9
CRVParadigm=Circuit-based2026.056362.8
Option ShufflingParadigm=Counterfactual-based2026.056116.7
Answer TracingParadigm=Logits-based2026.0553.230.8
PerplexityParadigm=Baselines2026.0549.436.1
RandomParadigm=Baselines2026.0543.237.4
Information GainParadigm=Logits-based2026.0541.620.2