List your company

CoT faithfulness detection on HLE Bio

78AccuracyCIE-SCORER
34.00845.42956.8568.271May 25, 2026

Evaluation results

MethodLinks
CIE-SCORERParadigm=Circuit-based2026.057879.7
BiGGenParadigm=LLM-as-judge2026.0570.269.2
Answer TracingParadigm=Logits-based2026.0564.376.2
Information GainParadigm=Logits-based2026.0552.59.5
PerplexityParadigm=Baselines2026.0547.452.4
Adding MistakesParadigm=Counterfactual-based2026.0546.451.6
Early AnsweringParadigm=Counterfactual-based2026.0546.448.3
ParaphrasingParadigm=Counterfactual-based2026.0545.740
Option ShufflingParadigm=Counterfactual-based2026.054014.3
Removing StepsParadigm=Counterfactual-based2026.0539.337
RandomParadigm=Baselines2026.0535.743.8