Long-form Question Answering on AlpacaFact MixP (50% polluted)
81.7VeriScore F1@kMIRAGE
Evaluation Results
| Method | Links | |
|---|---|---|
| MIRAGEBackbone=GPT-4o-mini2026.07 | 81.7 | |
| No-RAGBackbone=GPT-4o-mini2026.07 | 76.41 | |
| MIRAGEBackbone=Qwen3-8B2026.07 | 72.34 | |
| SelfRAGBackbone=GPT-4o-mini2026.07 | 70.38 | |
| RetRobust-NLIBackbone=GPT-4o-mini2026.07 | 69.97 | |
| AstuteRAGBackbone=GPT-4o-mini2026.07 | 68.11 | |
| DRAGINBackbone=GPT-4o-mini2026.07 | 67.43 | |
| AstuteRAGBackbone=Qwen3-8B2026.07 | 65.73 | |
| No-RAGBackbone=Qwen3-8B2026.07 | 65.71 | |
| BRIDGEBackbone=GPT-4o-mini2026.07 | 64.98 | |
| SelfRAGBackbone=Qwen3-8B2026.07 | 63.57 | |
| BRIDGEBackbone=Qwen3-8B2026.07 | 61.42 | |
| RobustRAGBackbone=Qwen3-8B2026.07 | 61.12 | |
| TrustRAGBackbone=GPT-4o-mini2026.07 | 60.62 | |
| FLAREBackbone=GPT-4o-mini2026.07 | 60.53 | |
| MADAMRAGBackbone=GPT-4o-mini2026.07 | 59.59 | |
| RetRobust-NLIBackbone=Qwen3-8B2026.07 | 59.13 | |
| RAGBackbone=GPT-4o-mini2026.07 | 58.92 | |
| RARGBackbone=GPT-4o-mini2026.07 | 57.32 | |
| DRAGINBackbone=Qwen3-8B2026.07 | 57.1 | |
| FLAREBackbone=Qwen3-8B2026.07 | 55.34 | |
| TrustRAGBackbone=Qwen3-8B2026.07 | 55.26 | |
| RAGBackbone=Qwen3-8B2026.07 | 55.15 | |
| RARGBackbone=Qwen3-8B2026.07 | 53.88 | |
| MADAMRAGBackbone=Qwen3-8B2026.07 | 53.61 | |
| RobustRAGBackbone=GPT-4o-mini2026.07 | 45.81 |