Factuality Hallucination on LongFact
23.5Facts ScoreNuggets
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| NuggetsBackbone=Qwen-2, Selection Ratio=10%2025.02 | 23.5 | 85.2 | 79.7 | 82.5 | |
| NuggetsBackbone=Qwen-2, Selection Ratio=15%2025.02 | 21.7 | 80.8 | 80.1 | 80.5 | |
| NOVABackbone=LLaMA-1, Selection Ratio=5%2025.02 | 21.5 | 88.1 | 82.5 | 85.3 | |
| CaRBackbone=LLaMA-1, Selection Ratio=10%2025.02 | 21.1 | 83.4 | 79.2 | 81.3 | |
| CaRBackbone=Qwen-2, Selection Ratio=15%2025.02 | 20.8 | 81.2 | 81.3 | 81.3 | |
| NOVABackbone=LLaMA-1, Selection Ratio=10%2025.02 | 20.7 | 87.6 | 83.2 | 85.4 | |
| CaRBackbone=Qwen-2, Selection Ratio=10%2025.02 | 20.3 | 84.5 | 79.8 | 82.2 | |
| IFDBackbone=LLaMA-1, Selection Ratio=5%2025.02 | 20.1 | 83.2 | 80.4 | 81.8 | |
| NuggetsBackbone=LLaMA-1, Selection Ratio=5%2025.02 | 19.4 | 85.1 | 77.3 | 81.2 | |
| NOVABackbone=LLaMA-1, Selection Ratio=15%2025.02 | 19.4 | 89.6 | 83.7 | 86.7 | |
| CaRBackbone=LLaMA-1, Selection Ratio=15%2025.02 | 19.3 | 84.2 | 80.4 | 82.3 | |
| NOVABackbone=Qwen-2, Selection Ratio=15%2025.02 | 19.3 | 88.8 | 82.9 | 85.9 | |
| IFDBackbone=LLaMA-1, Selection Ratio=10%2025.02 | 19.2 | 85.2 | 80.3 | 82.8 | |
| IFDBackbone=Qwen-2, Selection Ratio=15%2025.02 | 19.2 | 82.5 | 79.5 | 81 | |
| NuggetsBackbone=LLaMA-1, Selection Ratio=10%2025.02 | 18.8 | 84.2 | 78.6 | 81.4 | |
| IFDBackbone=LLaMA-1, Selection Ratio=15%2025.02 | 18.7 | 86.1 | 81.1 | 83.6 | |
| CaRBackbone=Qwen-2, Selection Ratio=5%2025.02 | 18.6 | 84.1 | 81.5 | 82.8 | |
| NOVABackbone=Qwen-2, Selection Ratio=10%2025.02 | 18.6 | 89.6 | 83.5 | 86.6 | |
| NOVABackbone=Qwen-2, Selection Ratio=5%2025.02 | 18.3 | 90.2 | 83.2 | 86.7 | |
| CaRBackbone=LLaMA-1, Selection Ratio=5%2025.02 | 18.2 | 85.9 | 80.1 | 83 | |
| NuggetsBackbone=LLaMA-1, Selection Ratio=15%2025.02 | 18.1 | 83.3 | 80 | 81.7 | |
| NuggetsBackbone=Qwen-2, Selection Ratio=5%2025.02 | 17.8 | 84.7 | 81 | 82.9 | |
| IFDBackbone=Qwen-2, Selection Ratio=10%2025.02 | 17.8 | 84.2 | 80.5 | 82.4 | |
| VanillaBackbone=Qwen-2, Selection Ratio=100%2025.02 | 17.3 | 83.4 | 80.2 | 81.8 | |
| FLAME-DPOBackbone=Qwen-2, Selection Ratio=100%2025.02 | 16.9 | 87.8 | 82.7 | 85.3 | |
| VanillaBackbone=LLaMA-1, Selection Ratio=100%2025.02 | 16.6 | 84.3 | 78.2 | 81.3 | |
| IFDBackbone=Qwen-2, Selection Ratio=5%2025.02 | 16.5 | 85.2 | 81.2 | 83.2 | |
| SELF-EVALBackbone=LLaMA-1, Selection Ratio=100%2025.02 | 15.7 | 87 | 80.8 | 83.9 | |
| FLAME-DPOBackbone=LLaMA-1, Selection Ratio=100%2025.02 | 14.8 | 86.7 | 81.2 | 84 | |
| SELF-EVALBackbone=Qwen-2, Selection Ratio=100%2025.02 | 14.2 | 88.2 | 81.6 | 84.9 |