List your company

Token-level hallucination detection on RAGTruth

59.4AP (Token-level)CORTEX
2.574417.327232.0846.8328Jun 30, 2026

Evaluation results

MethodLinks
CORTEXBackbone=Qwen2026.0659.493.72
CORTEXBackbone=Llama2026.0656.8692.75
CORTEXBackbone=Mistral2026.0654.7392.44
RAGLensBackbone=Qwen2026.0645.7590.75
RAGLensBackbone=Llama2026.0642.7189.53
RAGLensBackbone=Mistral2026.0642.4488.84
SAPLMABackbone=Qwen2026.0641.8288.2
SAPLMABackbone=Mistral2026.0640.2788.22
SAPLMABackbone=Llama2026.0638.9488.79
ICR ProbeBackbone=Llama2026.0626.6553.8
ICR ProbeBackbone=Mistral2026.0626.2153.11
ICR ProbeBackbone=Qwen2026.0624.6450.67
LLM-CheckBackbone=Qwen2026.069.5164.8
LLM-CheckBackbone=Llama2026.067.6960.79
LLM-CheckBackbone=Mistral2026.067.4361.89
NLLBackbone=Qwen2026.064.9546.86
NLLBackbone=Mistral2026.064.7946.66
NLLBackbone=Llama2026.064.7646.76