Robust Question Answering on RQA Evolving evidence streams GPT-4o (test)
72.68AccuracyInstructRAG
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| InstructRAGAttack=Benign, Pos=-2026.05 | 72.68 | — | |
| RADARAttack=Benign, Pos=-2026.05 | 61.29 | — | |
| RADARAttack=PIA, Pos=Pos 502026.05 | 61.23 | 10.87 | |
| RADARAttack=PIA, Pos=Pos 252026.05 | 60.65 | 12.15 | |
| AstuteRAGAttack=Benign, Pos=-2026.05 | 58.93 | — | |
| ReliabilityRAGAttack=Benign, Pos=-2026.05 | 58.67 | — | |
| RobustRAGAttack=Benign, Pos=-2026.05 | 58.46 | — | |
| ReliabilityRAGAttack=PIA, Pos=Pos 502026.05 | 58.09 | 4.54 | |
| ReliabilityRAGAttack=PIA, Pos=Pos 252026.05 | 56.88 | 8.25 | |
| RADARAttack=PIA, Pos=Pos 12026.05 | 56.17 | 20.02 | |
| Vanilla RAGAttack=Benign, Pos=-2026.05 | 56.05 | — | |
| InstructRAGAttack=PIA, Pos=Pos 12026.05 | 55.39 | 12.8 | |
| InstructRAGAttack=PIA, Pos=Pos 252026.05 | 53.93 | 32.18 | |
| RobustRAGAttack=PIA, Pos=Pos 252026.05 | 53.49 | 26.49 | |
| InstructRAGAttack=PIA, Pos=Pos 502026.05 | 53.42 | 31.35 | |
| RobustRAGAttack=PIA, Pos=Pos 502026.05 | 53.29 | 25.14 | |
| AstuteRAGAttack=PIA, Pos=Pos 502026.05 | 50.99 | 10.68 | |
| AstuteRAGAttack=PIA, Pos=Pos 252026.05 | 50.93 | 10.75 | |
| RobustRAGAttack=PIA, Pos=Pos 12026.05 | 47.47 | 31.99 | |
| AstuteRAGAttack=PIA, Pos=Pos 12026.05 | 47.28 | 10.81 | |
| Vanilla RAGAttack=PIA, Pos=Pos 252026.05 | 45.04 | 25.4 | |
| ReliabilityRAGAttack=PIA, Pos=Pos 12026.05 | 43.38 | 39.41 | |
| Vanilla RAGAttack=PIA, Pos=Pos 502026.05 | 40.12 | 27.9 | |
| Vanilla RAGAttack=PIA, Pos=Pos 12026.05 | 35.76 | 39.99 |