Token-level hallucination detection on HalluRAG
76.9APCORTEX
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CORTEXBackbone=Llama2026.06 | 76.9 | 83.6 | |
| CORTEXBackbone=Qwen2026.06 | 74.95 | 84.26 | |
| CORTEXBackbone=Mistral2026.06 | 73.08 | 82.33 | |
| SAPLMABackbone=Llama2026.06 | 68.75 | 80.24 | |
| SAPLMABackbone=Qwen2026.06 | 65.84 | 81.29 | |
| RAGLensBackbone=Qwen2026.06 | 63.8 | 76.39 | |
| SAPLMABackbone=Mistral2026.06 | 61.93 | 75.78 | |
| RAGLensBackbone=Llama2026.06 | 57.25 | 74.25 | |
| RAGLensBackbone=Mistral2026.06 | 55.96 | 73.93 | |
| ICR ProbeBackbone=Llama2026.06 | 32.92 | 52.85 | |
| ICR ProbeBackbone=Mistral2026.06 | 31.97 | 51.43 | |
| ICR ProbeBackbone=Qwen2026.06 | 31.31 | 51.06 | |
| LLM-CheckBackbone=Llama2026.06 | 29.94 | 54.11 | |
| LLM-CheckBackbone=Mistral2026.06 | 29.44 | 54.67 | |
| LLM-CheckBackbone=Qwen2026.06 | 28.74 | 53.42 | |
| NLLBackbone=Mistral2026.06 | 26.85 | 49.45 | |
| NLLBackbone=Qwen2026.06 | 26.58 | 49.34 | |
| NLLBackbone=Llama2026.06 | 26.48 | 49.52 |