List your company
ResearchFormerly WizWand

PMI ranking estimation on ChaosNLI (500 held-out pairs)

0.82Spearman RhoPromptNCE
0.47680.56590.6550.7441May 20, 2026
Updated 5mo ago

Evaluation results

MethodLinks
PromptNCEModel=Claude Sonnet 4, Evaluation Type=Zero-shot2026.050.82
PromptNCE (emp.)Model=Claude Sonnet 4, Evaluation Type=Empirical marginal (oracle upper bound)2026.050.81
Decomposed PMIModel=Claude Sonnet 4, Evaluation Type=Zero-shot2026.050.75
InfoNCE (emp.)Model=Claude Sonnet 4, Evaluation Type=Empirical marginal (oracle upper bound)2026.050.73
PromptNCE (emp.)Model=GPT-5.2, Evaluation Type=Empirical marginal (oracle upper bound)2026.050.73
InfoNCEModel=Claude Sonnet 4, Evaluation Type=Zero-shot2026.050.73
MarginalNCEModel=Claude Sonnet 4, Evaluation Type=Zero-shot2026.050.73
PromptNCEModel=GPT-5.2, Evaluation Type=Zero-shot2026.050.73
Direct PMIModel=Claude Sonnet 4, Evaluation Type=Zero-shot2026.050.72
Decomposed PMI (emp.)Model=Claude Sonnet 4, Evaluation Type=Empirical marginal (oracle upper bound)2026.050.7
InfoNCE (emp.)Model=GPT-5.2, Evaluation Type=Empirical marginal (oracle upper bound)2026.050.7
Decomposed PMIModel=GPT-5.2, Evaluation Type=Zero-shot2026.050.7
Decomposed PMI (emp.)Model=GPT-5.2, Evaluation Type=Empirical marginal (oracle upper bound)2026.050.69
InfoNCEModel=GPT-5.2, Evaluation Type=Zero-shot2026.050.69
MarginalNCEModel=GPT-5.2, Evaluation Type=Zero-shot2026.050.69
Direct PMIModel=GPT-5.2, Evaluation Type=Zero-shot2026.050.49