List your company

Hallucination Detection on MNLI (test)

100AuROCMarginal Em
24.0843.7963.583.21Feb 21, 2026

Evaluation results

MethodLinks
Marginal EmModel=Gemma-Instruct 4B, Pool=Max2026.02100
Spilled ΔEModel=Gemma-Instruct 4B, Pool=Min2026.0299.66
Marginal EmModel=Gemma-Instruct 4B, Pool=Min2026.0299.01
Marginal EmModel=Gemma-Instruct 1B, Pool=Max2026.0298.17
Spilled ΔEsModel=Gemma-Instruct 4B, Pool=Max2026.0295.61
Spilled ΔEModel=Gemma-Instruct 1B, Pool=Min2026.0292.93
Spilled ΔEsModel=Gemma-Instruct 1B, Pool=Max2026.0277.48
Marginal EmModel=Gemma-Instruct 1B, Pool=Min2026.0256.33
Logit EℓModel=Gemma-Instruct 1B, Pool=Max2026.0249.28
Logit EℓModel=Gemma-Instruct 4B, Pool=Max2026.0227