Answer-level hallucination detection on HalluRAG
80.1APCORTEX
Evaluation results
| Method | Links | ||
|---|---|---|---|
| CORTEXBackbone LLM=Mistral2026.06 | 80.1 | 87.18 | |
| RAGLensBackbone LLM=Mistral2026.06 | 80.06 | 88.35 | |
| RAGLensBackbone LLM=Llama2026.06 | 79.94 | 89.05 | |
| SAPLMABackbone LLM=Llama2026.06 | 79.22 | 84.82 | |
| CORTEXBackbone LLM=Qwen2026.06 | 79.07 | 86.27 | |
| SAPLMABackbone LLM=Qwen2026.06 | 78.67 | 86.95 | |
| CORTEXBackbone LLM=Llama2026.06 | 77.95 | 84.81 | |
| RAGLensBackbone LLM=Qwen2026.06 | 76.69 | 83.8 | |
| SAPLMABackbone LLM=Mistral2026.06 | 75.29 | 86.28 | |
| ICR ProbeBackbone LLM=Llama2026.06 | 40.77 | 66.4 | |
| LLM-CheckBackbone LLM=Mistral2026.06 | 40.59 | 64.81 | |
| LLM-CheckBackbone LLM=Qwen2026.06 | 39.56 | 66.38 | |
| ICR ProbeBackbone LLM=Qwen2026.06 | 32.86 | 60.31 | |
| NLLBackbone LLM=Qwen2026.06 | 31.84 | 58.06 | |
| LLM-CheckBackbone LLM=Llama2026.06 | 30.52 | 57.11 | |
| NLLBackbone LLM=Mistral2026.06 | 29.69 | 56 | |
| ICR ProbeBackbone LLM=Mistral2026.06 | 21.57 | 43.1 | |
| NLLBackbone LLM=Llama2026.06 | 20.55 | 48.27 |