List your company

Hallucination Detection on HotpotQA-WC (test)

0.8595AuROCMarginal Em
0.3376280.4731140.60860.744086Feb 21, 2026

Evaluation results

MethodLinks
Marginal EmModel=Gemma-Instruct 4B, Pool=Max2026.020.8595
Spilled ΔEModel=Gemma-Instruct 4B, Pool=Min2026.020.827
Marginal EmModel=Gemma-Instruct 1B, Pool=Max2026.020.8233
Logit EℓModel=Gemma-Instruct 1B, Pool=Max2026.020.7567
Spilled ΔEModel=Gemma-Instruct 1B, Pool=Min2026.020.7186
Marginal EmModel=Gemma-Instruct 4B, Pool=Min2026.020.7184
Logit EℓModel=Gemma-Instruct 4B, Pool=Max2026.020.6449
Marginal EmModel=Gemma-Instruct 1B, Pool=Min2026.020.6156
Spilled ΔEsModel=Gemma-Instruct 4B, Pool=Max2026.020.5121
Spilled ΔEsModel=Gemma-Instruct 1B, Pool=Max2026.020.3577