List your company

Discrimination Test on Reference Set (N=500)

100Accept (valid)CoT-Instructed
9597.5100102.5Dec 16, 2025

Evaluation results

MethodLinks
CoT-InstructedBase Model=GPT-5.1, Method Category=Outcome-Based (Self-Correction)2025.12100——
Self-ConsistencyBase Model=GPT-5.1, Method Category=Outcome-Based (Self-Correction)2025.12100——
ReflexionBase Model=GPT-5.1, Method Category=Outcome-Based (Self-Correction)2025.12100——
Self-RefineBase Model=GPT-5.1, Method Category=Outcome-Based (Self-Correction)2025.12100——
RCABase Model=GPT-5.1, Method Category=Process-Based (Ours)2025.12——0