List your company

Question Answering on DROP (Pearson r)

0.89Pearson rCoT-SC
0.18280.36640.550.7336Jun 11, 2026

Evaluation results

MethodLinks
CoT-SCK=3, Format=continuous2026.060.89
CoT-SCK=10, Format=continuous2026.060.87
OCFormat=binarized at 0.5, Signal=paper’s primary2026.060.86
OCFormat=continuous2026.060.85
Skywork-PRMMechanism=min over steps2026.060.82
naive direct-answer SCK=3, Mechanism=ablation2026.060.62
−HK=10, Format=continuous2026.060.61
naive direct-answer SCK=10, Mechanism=ablation2026.060.61
RLHFlow-PRMMechanism=min over steps2026.060.31
P(True)Format=continuous, Models=9 models2026.060.21
decomposed-SCK=102026.060.21