Question Answering on TQA poison @ Position 1, k=10 (test)
66.4Robustness AccuracyAstuteRAG
Evaluation Results
| Method | Links | |
|---|---|---|
| AstuteRAGModel=GPT-4o-mini, k=102025.09 | 66.4 | |
| RobustRAGModel=GPT-4o-mini, k=102025.09 | 60.4 | |
| MISModel=GPT-4o-mini, k=102025.09 | 60.4 | |
| RobustRAGModel=Llama3.2-3B, k=102025.09 | 60 | |
| MISModel=Mistral-7B, k=102025.09 | 59 | |
| MISModel=Llama3.2-3B, k=102025.09 | 58.6 | |
| RobustRAGModel=Mistral-7B, k=102025.09 | 56.8 | |
| AstuteRAGModel=Mistral-7B, k=102025.09 | 54.2 | |
| AstuteRAGModel=Llama3.2-3B, k=102025.09 | 53.2 | |
| InstructRAGModel=Mistral-7B, k=102025.09 | 49.2 | |
| Vanilla RAGModel=Mistral-7B, k=102025.09 | 46 | |
| InstructRAGModel=GPT-4o-mini, k=102025.09 | 44.8 | |
| Vanilla RAGModel=Llama3.2-3B, k=102025.09 | 43.6 | |
| Vanilla RAGModel=GPT-4o-mini, k=102025.09 | 43.4 | |
| InstructRAGModel=Llama3.2-3B, k=102025.09 | 32.4 |