Question Answering on TQA PIA Attack (test)
76.4AccuracyReliabilityRAG
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ReliabilityRAGPos=25, top-k=502026.05 | 76.4 | 7.6 | |
| ReliabilityRAGPos=50, top-k=502026.05 | 76.4 | 5.4 | |
| RADARPos=50, top-k=502026.05 | 70.6 | 16 | |
| RADARPos=25, top-k=502026.05 | 70.4 | 16.2 | |
| RobustRAGPos=25, top-k=502026.05 | 68.2 | 11.6 | |
| RobustRAGPos=50, top-k=502026.05 | 68 | 11.4 | |
| RADARPos=1, top-k=502026.05 | 64.6 | 26.2 | |
| InstructRAGPos=50, top-k=502026.05 | 63.8 | 5.4 | |
| InstructRAGPos=25, top-k=502026.05 | 63.6 | 4.6 | |
| Vanilla RAGPos=25, top-k=502026.05 | 62.8 | 33.8 | |
| Vanilla RAGPos=50, top-k=502026.05 | 62.6 | 30.2 | |
| RobustRAGPos=1, top-k=502026.05 | 62 | 21.6 | |
| ReliabilityRAGPos=1, top-k=502026.05 | 58.8 | 36 | |
| AstuteRAGPos=50, top-k=502026.05 | 43.8 | 2 | |
| AstuteRAGPos=25, top-k=502026.05 | 43 | 1.8 | |
| AstuteRAGPos=1, top-k=502026.05 | 39 | 5.6 | |
| InstructRAGPos=1, top-k=502026.05 | 31 | 58 | |
| Vanilla RAGPos=1, top-k=502026.05 | 12.4 | 88.8 |