Long-form Question Answering on LongFact MixP (50% polluted)
90.33VeriScore F1@kMIRAGE
Evaluation Results
| Method | Links | |
|---|---|---|
| MIRAGEBackbone=GPT-4o-mini2026.07 | 90.33 | |
| MIRAGEBackbone=Qwen3-8B2026.07 | 87.56 | |
| No-RAGBackbone=GPT-4o-mini2026.07 | 86.9 | |
| No-RAGBackbone=Qwen3-8B2026.07 | 84.61 | |
| AstuteRAGBackbone=GPT-4o-mini2026.07 | 81.14 | |
| BRIDGEBackbone=GPT-4o-mini2026.07 | 80.71 | |
| DRAGINBackbone=GPT-4o-mini2026.07 | 73.41 | |
| FLAREBackbone=GPT-4o-mini2026.07 | 73.32 | |
| SelfRAGBackbone=GPT-4o-mini2026.07 | 73.08 | |
| DRAGINBackbone=Qwen3-8B2026.07 | 71.48 | |
| BRIDGEBackbone=Qwen3-8B2026.07 | 70.5 | |
| RetRobust-NLIBackbone=GPT-4o-mini2026.07 | 70.27 | |
| RobustRAGBackbone=Qwen3-8B2026.07 | 69.21 | |
| AstuteRAGBackbone=Qwen3-8B2026.07 | 67.22 | |
| RobustRAGBackbone=GPT-4o-mini2026.07 | 66.22 | |
| TrustRAGBackbone=GPT-4o-mini2026.07 | 65.66 | |
| MADAMRAGBackbone=GPT-4o-mini2026.07 | 65.48 | |
| RARGBackbone=GPT-4o-mini2026.07 | 65.38 | |
| SelfRAGBackbone=Qwen3-8B2026.07 | 65.32 | |
| RetRobust-NLIBackbone=Qwen3-8B2026.07 | 65.15 | |
| MADAMRAGBackbone=Qwen3-8B2026.07 | 64.92 | |
| FLAREBackbone=Qwen3-8B2026.07 | 61.13 | |
| TrustRAGBackbone=Qwen3-8B2026.07 | 59.01 | |
| RAGBackbone=GPT-4o-mini2026.07 | 58.24 | |
| RAGBackbone=Qwen3-8B2026.07 | 57.13 | |
| RARGBackbone=Qwen3-8B2026.07 | 56.37 |