Long-form Question Answering on FAVA FullP (100% polluted)
82.54VeriScore F1@kMIRAGE
Evaluation Results
| Method | Links | |
|---|---|---|
| MIRAGEBackbone=GPT-4o-mini2026.07 | 82.54 | |
| No-RAGBackbone=GPT-4o-mini2026.07 | 81.2 | |
| MIRAGEBackbone=Qwen3-8B2026.07 | 77.35 | |
| No-RAGBackbone=Qwen3-8B2026.07 | 74.92 | |
| AstuteRAGBackbone=GPT-4o-mini2026.07 | 72.87 | |
| BRIDGEBackbone=GPT-4o-mini2026.07 | 67.18 | |
| RetRobust-NLIBackbone=GPT-4o-mini2026.07 | 64.26 | |
| FLAREBackbone=GPT-4o-mini2026.07 | 63.89 | |
| DRAGINBackbone=GPT-4o-mini2026.07 | 61.25 | |
| AstuteRAGBackbone=Qwen3-8B2026.07 | 54.66 | |
| BRIDGEBackbone=Qwen3-8B2026.07 | 49.94 | |
| SelfRAGBackbone=Qwen3-8B2026.07 | 46.69 | |
| FLAREBackbone=Qwen3-8B2026.07 | 45.33 | |
| TrustRAGBackbone=GPT-4o-mini2026.07 | 42.64 | |
| SelfRAGBackbone=GPT-4o-mini2026.07 | 42.44 | |
| MADAMRAGBackbone=GPT-4o-mini2026.07 | 42.25 | |
| MADAMRAGBackbone=Qwen3-8B2026.07 | 41.98 | |
| RARGBackbone=GPT-4o-mini2026.07 | 41.7 | |
| RetRobust-NLIBackbone=Qwen3-8B2026.07 | 40.4 | |
| DRAGINBackbone=Qwen3-8B2026.07 | 36.55 | |
| RobustRAGBackbone=Qwen3-8B2026.07 | 35.27 | |
| RAGBackbone=GPT-4o-mini2026.07 | 34.83 | |
| TrustRAGBackbone=Qwen3-8B2026.07 | 32.63 | |
| RARGBackbone=Qwen3-8B2026.07 | 32.25 | |
| RAGBackbone=Qwen3-8B2026.07 | 31.28 | |
| RobustRAGBackbone=GPT-4o-mini2026.07 | 26.98 |