Hallucination Detection on RAGTruth (Consistency Rates)
92.67Summary Consistency RateMARCH-General
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| MARCH-GeneralBackbone=Meta-Llama3.1-8B-Instruct2026.03 | 92.67 | 72 | 83.45 | |
| GLM-4-9B-ChatParameters=9B2026.03 | 87.77 | 52.67 | 90.67 | |
| MARCH-STEMBackbone=Meta-Llama3.1-8B-Instruct2026.03 | 86.67 | 72.67 | 83.45 | |
| Ministral-8B-InstructParameters=8B2026.03 | 83.33 | 49.33 | 88.49 | |
| Qwen2.5-14B-InstructParameters=14B2026.03 | 82 | 51.33 | 93.53 | |
| Phi-4-mini-Instruct2026.03 | 78.67 | 24.67 | 94.96 | |
| o4-mini-high-2025-04-162026.03 | 76.67 | 64 | 82.01 | |
| o4-mini-medium-2025-04-162026.03 | 76.67 | 67.33 | 88.49 | |
| Meta-Llama3.1-8B-InstructParameters=8B2026.03 | 71.33 | 48.67 | 63.31 | |
| o4-mini-low-2025-04-162026.03 | 70 | 71.33 | 83.45 |