Dynamic Retrieval-Augmented Generation on Bio (test)
83.1AccuracyInstructRAG
Evaluation results
| Method | Links | ||
|---|---|---|---|
| InstructRAGAttack=Benign, Pos=–, History Setting=lightweight history, LLM=DeepSeek, top-k=502026.05 | 83.1 | — | |
| RADARAttack=Benign, Pos=–, History Setting=lightweight history, LLM=DeepSeek, top-k=502026.05 | 74.02 | — | |
| ReliabilityRAGAttack=Benign, Pos=–, History Setting=lightweight history, LLM=DeepSeek, top-k=502026.05 | 72.55 | — | |
| Vanilla RAGAttack=Benign, Pos=–, History Setting=lightweight history, LLM=DeepSeek, top-k=502026.05 | 70.7 | — | |
| RADARAttack=PIA, Pos=Pos 25, History Setting=lightweight history, LLM=DeepSeek, top-k=502026.05 | 70.12 | 8.94 | |
| RADARAttack=PIA, Pos=Pos 50, History Setting=lightweight history, LLM=DeepSeek, top-k=502026.05 | 70.05 | 6.01 | |
| ReliabilityRAGAttack=PIA, Pos=Pos 50, History Setting=lightweight history, LLM=DeepSeek, top-k=502026.05 | 68.45 | 6.78 | |
| ReliabilityRAGAttack=PIA, Pos=Pos 25, History Setting=lightweight history, LLM=DeepSeek, top-k=502026.05 | 66.53 | 10.81 | |
| AstuteRAGAttack=PIA, Pos=Pos 25, History Setting=lightweight history, LLM=DeepSeek, top-k=502026.05 | 64.68 | 8.25 | |
| AstuteRAGAttack=Benign, Pos=–, History Setting=lightweight history, LLM=DeepSeek, top-k=502026.05 | 64.17 | — | |
| RADARAttack=PIA, Pos=Pos 1, History Setting=lightweight history, LLM=DeepSeek, top-k=502026.05 | 63.6 | 17.85 | |
| AstuteRAGAttack=PIA, Pos=Pos 50, History Setting=lightweight history, LLM=DeepSeek, top-k=502026.05 | 62.06 | 9.85 | |
| AstuteRAGAttack=PIA, Pos=Pos 1, History Setting=lightweight history, LLM=DeepSeek, top-k=502026.05 | 60.08 | 7.74 | |
| RobustRAGAttack=Benign, Pos=–, History Setting=lightweight history, LLM=DeepSeek, top-k=502026.05 | 59.88 | — | |
| RobustRAGAttack=PIA, Pos=Pos 25, History Setting=lightweight history, LLM=DeepSeek, top-k=502026.05 | 59.31 | 12.92 | |
| RobustRAGAttack=PIA, Pos=Pos 50, History Setting=lightweight history, LLM=DeepSeek, top-k=502026.05 | 59.12 | 12.8 | |
| InstructRAGAttack=PIA, Pos=Pos 1, History Setting=lightweight history, LLM=DeepSeek, top-k=502026.05 | 57.38 | 32.69 | |
| InstructRAGAttack=PIA, Pos=Pos 25, History Setting=lightweight history, LLM=DeepSeek, top-k=502026.05 | 55.34 | 25.46 | |
| RobustRAGAttack=PIA, Pos=Pos 1, History Setting=lightweight history, LLM=DeepSeek, top-k=502026.05 | 53.49 | 20.15 | |
| ReliabilityRAGAttack=PIA, Pos=Pos 1, History Setting=lightweight history, LLM=DeepSeek, top-k=502026.05 | 50.74 | 43.12 | |
| InstructRAGAttack=PIA, Pos=Pos 50, History Setting=lightweight history, LLM=DeepSeek, top-k=502026.05 | 42.22 | 49.07 | |
| Vanilla RAGAttack=PIA, Pos=Pos 1, History Setting=lightweight history, LLM=DeepSeek, top-k=502026.05 | 37.68 | 60.72 | |
| Vanilla RAGAttack=PIA, Pos=Pos 25, History Setting=lightweight history, LLM=DeepSeek, top-k=502026.05 | 36.98 | 58.35 | |
| Vanilla RAGAttack=PIA, Pos=Pos 50, History Setting=lightweight history, LLM=DeepSeek, top-k=502026.05 | 14.84 | 82.47 |