Faithfulness Hallucination Detection on LLM-AggreFact & HoVer Refined
4.6Overall Std DevFaithLens
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| FaithLensbackbone=Llama-3.1-8B-Inst2025.12 | 4.6 | 86.4 | |
| DeepSeek-V3.2thinking=true2025.12 | 5.1 | 84.4 | |
| Claude-3.7-Sonnet2025.12 | 5.3 | 82.6 | |
| o12025.12 | 5.9 | 76.8 | |
| o3-mini2025.12 | 5.9 | 79.2 | |
| o32025.12 | 6 | 82.1 | |
| Llama-3.1-405B-Inst2025.12 | 6.4 | 75.8 | |
| GPT-4.12025.12 | 6.5 | 83 | |
| ClearCheckbackbone=Llama-3.1-8B-Inst2025.12 | 6.6 | 80.1 | |
| GPT-4o2025.12 | 7 | 76.1 | |
| FactCGparameters=435M2025.12 | 7 | 78.2 | |
| MiniCheckparameters=7B2025.12 | 7.5 | 80.7 | |
| DeepSeek-V3.2thinking=false2025.12 | 7.8 | 76.8 | |
| AlignScoreparameters=355M2025.12 | 12 | 75.6 |