Long-form Question Answering on Biography MixP (50% polluted)
70.11VeriScore F1@kMIRAGE
Evaluation Results
| Method | Links | |
|---|---|---|
| MIRAGEBackbone=GPT-4o-mini2026.07 | 70.11 | |
| MIRAGEBackbone=Qwen3-8B2026.07 | 61.25 | |
| No-RAGBackbone=GPT-4o-mini2026.07 | 58.99 | |
| MADAMRAGBackbone=GPT-4o-mini2026.07 | 58.39 | |
| AstuteRAGBackbone=GPT-4o-mini2026.07 | 56.55 | |
| BRIDGEBackbone=GPT-4o-mini2026.07 | 56.37 | |
| SelfRAGBackbone=GPT-4o-mini2026.07 | 56.23 | |
| No-RAGBackbone=Qwen3-8B2026.07 | 53.75 | |
| DRAGINBackbone=GPT-4o-mini2026.07 | 52.58 | |
| FLAREBackbone=GPT-4o-mini2026.07 | 51.58 | |
| RetRobust-NLIBackbone=GPT-4o-mini2026.07 | 49.5 | |
| MADAMRAGBackbone=Qwen3-8B2026.07 | 48.91 | |
| RARGBackbone=GPT-4o-mini2026.07 | 48.76 | |
| RobustRAGBackbone=Qwen3-8B2026.07 | 48.14 | |
| TrustRAGBackbone=GPT-4o-mini2026.07 | 48.03 | |
| RAGBackbone=GPT-4o-mini2026.07 | 42.56 | |
| RobustRAGBackbone=GPT-4o-mini2026.07 | 40.47 | |
| FLAREBackbone=Qwen3-8B2026.07 | 36.77 | |
| SelfRAGBackbone=Qwen3-8B2026.07 | 35.18 | |
| RARGBackbone=Qwen3-8B2026.07 | 34.03 | |
| TrustRAGBackbone=Qwen3-8B2026.07 | 33.64 | |
| DRAGINBackbone=Qwen3-8B2026.07 | 33.36 | |
| BRIDGEBackbone=Qwen3-8B2026.07 | 31.44 | |
| AstuteRAGBackbone=Qwen3-8B2026.07 | 28.86 | |
| RAGBackbone=Qwen3-8B2026.07 | 26.64 | |
| RetRobust-NLIBackbone=Qwen3-8B2026.07 | 15.27 |