List your company
ResearchFormerly WizWand

Hallucination Detection on RAGBench

67.9AUCGASP+base (trained)
56.87659.73862.665.462Jul 5, 2026
Updated 3mo ago

Evaluation results

MethodLinks
GASP+base (trained)Scorer=Qwen-1.5B2026.0767.9
GASP+base (trained)Scorer=Qwen-0.5B2026.0766.8
PerplexityScorer=Qwen-1.5B2026.0761.9
PerplexityScorer=Qwen-0.5B2026.0761.3
GASP-trainedScorer=Qwen-1.5B2026.0759.6
GASP-trainedScorer=Qwen-0.5B2026.0759.5
GASP-thresholdScorer=Qwen-0.5B2026.0757.6
GASP-thresholdScorer=Qwen-1.5B2026.0757.3