Question Answering on NQ (ACC_clean, ACC, ASR)
67Clean Accuracy (NQ)RobustRAG
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| RobustRAGBase Model=Mistral-7B, Variant=Keyword2026.05 | 67 | 66 | 9 | |
| Vanilla RAGBase Model=Vicuna-7B2026.05 | 67 | 31 | 45 | |
| TrustRAGBase Model=Mistral-7B2026.05 | 65 | 62 | 3 | |
| Vanilla RAGBase Model=Mistral-7B2026.05 | 61 | 26 | 46 | |
| PRA-RAGBase Model=Mistral-7B2026.05 | 60 | 62 | 1 | |
| RAGForensicsBase Model=Vicuna-7B2026.05 | 60 | 58 | 3 | |
| PRA-RAGBase Model=Mistral-7B, Variant=cat2026.05 | 57 | 47 | 9 | |
| PRA-RAGBase Model=Vicuna-7B, Variant=cat2026.05 | 54 | 53 | 4 | |
| PRA-RAGBase Model=Vicuna-7B2026.05 | 53 | 49 | 1 | |
| TrustRAGBase Model=Vicuna-7B2026.05 | 51 | 49 | 6 | |
| ASTUTE RAGBase Model=Vicuna-7B2026.05 | 50 | 44 | 19 | |
| ASTUTE RAGBase Model=Mistral-7B2026.05 | 48 | 47 | 5 | |
| RobustRAGBase Model=Vicuna-7B, Variant=Keyword2026.05 | 48 | 40 | 8 | |
| RAGForensicsBase Model=Mistral-7B2026.05 | 41 | 41 | 7 | |
| InstructRAGBase Model=Mistral-7B, Variant=ICL2026.05 | 38 | 26 | 13 | |
| InstructRAGBase Model=Vicuna-7B, Variant=ICL2026.05 | 33 | 10 | 11 |