List your company

Hallucination Detection on PubMedQA (AUROC)

94.8AUROCLatentAudit
92.40893.02993.6594.271Apr 7, 2026

Evaluation results

MethodLinks
LatentAuditModel Backbone=Qwen-3 (8B)2026.0494.8
LatentAuditModel Backbone=Llama-3 (8B)2026.0494.2
LatentAuditModel Backbone=Qwen-2.5 (7B)2026.0493.8
LatentAuditModel Backbone=Llama-2 (7B)2026.0493.1
LatentAuditModel Backbone=Mistral (7B)2026.0492.5