Retrieval-Augmented Generation on MS MARCO
90Accuracy (Clean)VanillaRAG
Evaluation Results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| VanillaRAGModel=Llama-3.1-8B, Retriever=ANCE2026.05 | 90 | — | — | — | — | — | — | — | 29 | 83 | 82 | 77 | |
| VanillaRAGModel=Llama-3.1-8B, Retriever=Contriever2026.05 | 90 | — | — | — | — | — | — | — | 35 | 82 | 89 | 74 | |
| VanillaRAGModel=Llama-3.1-8B, Retriever=DPR2026.05 | 89 | — | — | — | — | — | — | — | 31 | 73 | 81 | 79 | |
| VanillaRAGModel=Mistral-7B, Retriever=Contriever2026.05 | 86 | — | — | — | — | — | — | — | 20 | 59 | 75 | 32 | |
| BiRDModel=Llama-3.1-8B, Retriever=DPR, Venue=Ours2026.05 | 83 | — | — | — | — | — | — | — | 35 | 76 | 36 | 83 | |
| BiRDModel=Llama-3.1-8B, Retriever=ANCE, Venue=Ours2026.05 | 82 | — | — | — | — | — | — | — | 28 | 82 | 30 | 87 | |
| VanillaRAGModel=Qwen2.5-7B, Retriever=ANCE2026.05 | 81 | — | — | — | — | — | — | — | 26 | 69 | 86 | 47 | |
| VanillaRAGModel=Qwen2.5-7B, Retriever=Contriever2026.05 | 81 | — | — | — | — | — | — | — | 26 | 63 | 85 | 49 | |
| VanillaRAGModel=Mistral-7B, Retriever=ANCE2026.05 | 81 | — | — | — | — | — | — | — | 13 | 72 | 88 | 24 | |
| BiRDModel=Llama-3.1-8B, Retriever=Contriever, Venue=Ours2026.05 | 81 | — | — | — | — | — | — | — | 36 | 83 | 35 | 87 | |
| InstructRAGModel=Qwen2.5-7B, Retriever=ANCE, Venue=ICLR ’252026.05 | 79 | — | — | — | — | — | — | — | 63 | 56 | 80 | 35 | |
| VanillaRAGModel=Qwen2.5-7B, Retriever=DPR2026.05 | 78 | — | — | — | — | — | — | — | 26 | 65 | 88 | 47 | |
| VanillaRAGModel=Mistral-7B, Retriever=DPR2026.05 | 77 | — | — | — | — | — | — | — | 23 | 55 | 76 | 32 | |
| InstructRAGModel=Qwen2.5-7B, Retriever=Contriever, Venue=ICLR ’252026.05 | 76 | — | — | — | — | — | — | — | 53 | 55 | 84 | 35 | |
| InstructRAGModel=Llama-3.1-8B, Retriever=ANCE, Venue=ICLR ’252026.05 | 76 | — | — | — | — | — | — | — | 39 | 72 | 60 | 70 | |
| InstructRAGModel=Qwen2.5-7B, Retriever=DPR, Venue=ICLR ’252026.05 | 75 | — | — | — | — | — | — | — | 50 | 55 | 79 | 32 | |
| InstructRAGModel=Llama-3.1-8B, Retriever=Contriever, Venue=ICLR ’252026.05 | 74 | — | — | — | — | — | — | — | 37 | 74 | 58 | 74 | |
| InstructRAGModel=Llama-3.1-8B, Retriever=DPR, Venue=ICLR ’252026.05 | 74 | — | — | — | — | — | — | — | 49 | 75 | 61 | 69 | |
| RobustRAGModel=Qwen2.5-7B, Retriever=ANCE, Venue=arXiv ’242026.05 | 71 | — | — | — | — | — | — | — | 69 | 52 | 69 | 56 | |
| BiRDModel=Mistral-7B, Retriever=Contriever, Venue=Ours2026.05 | 71 | — | — | — | — | — | — | — | 23 | 62 | 14 | 76 | |
| ReliabilityRAGModel=Qwen2.5-7B, Retriever=DPR, Venue=NeurIPS ’252026.05 | 70 | — | — | — | — | — | — | — | 31 | 59 | 58 | 57 | |
| RobustRAGModel=Qwen2.5-7B, Retriever=DPR, Venue=arXiv ’242026.05 | 69 | — | — | — | — | — | — | — | 65 | 58 | 67 | 51 | |
| BiRDModel=Mistral-7B, Retriever=ANCE, Venue=Ours2026.05 | 69 | — | — | — | — | — | — | — | 12 | 69 | 12 | 75 | |
| BiRDModel=Mistral-7B, Retriever=DPR, Venue=Ours2026.05 | 67 | — | — | — | — | — | — | — | 20 | 61 | 14 | 71 | |
| RobustRAGModel=Qwen2.5-7B, Retriever=Contriever, Venue=arXiv ’242026.05 | 66 | — | — | — | — | — | — | — | 71 | 61 | 69 | 48 | |
| ReliabilityRAGModel=Qwen2.5-7B, Retriever=ANCE, Venue=NeurIPS ’252026.05 | 64 | — | — | — | — | — | — | — | 27 | 59 | 58 | 60 | |
| ReliabilityRAGModel=Qwen2.5-7B, Retriever=Contriever, Venue=NeurIPS ’252026.05 | 64 | — | — | — | — | — | — | — | 30 | 62 | 57 | 55 | |
| BiRDModel=Qwen2.5-7B, Retriever=DPR, Venue=Ours2026.05 | 64 | — | — | — | — | — | — | — | 25 | 63 | 27 | 68 | |
| BiRDModel=Qwen2.5-7B, Retriever=Contriever, Venue=Ours2026.05 | 60 | — | — | — | — | — | — | — | 27 | 63 | 24 | 71 | |
| BiRDModel=Qwen2.5-7B, Retriever=ANCE, Venue=Ours2026.05 | 59 | — | — | — | — | — | — | — | 22 | 69 | 17 | 76 | |
| ReliabilityRAGModel=Llama-3.1-8B, Retriever=ANCE, Venue=NeurIPS ’252026.05 | 40 | — | — | — | — | — | — | — | 67 | 48 | 46 | 44 | |
| RobustRAGModel=Llama-3.1-8B, Retriever=Contriever, Venue=arXiv ’242026.05 | 40 | — | — | — | — | — | — | — | 72 | 54 | 39 | 49 | |
| ReliabilityRAGModel=Llama-3.1-8B, Retriever=Contriever, Venue=NeurIPS ’252026.05 | 39 | — | — | — | — | — | — | — | 65 | 54 | 33 | 34 | |
| RobustRAGModel=Llama-3.1-8B, Retriever=DPR, Venue=arXiv ’242026.05 | 35 | — | — | — | — | — | — | — | 59 | 59 | 43 | 51 | |
| RobustRAGModel=Llama-3.1-8B, Retriever=ANCE, Venue=arXiv ’242026.05 | 34 | — | — | — | — | — | — | — | 81 | 61 | 39 | 50 | |
| ReliabilityRAGModel=Llama-3.1-8B, Retriever=DPR, Venue=NeurIPS ’252026.05 | 30 | — | — | — | — | — | — | — | 56 | 53 | 37 | 39 | |
| InstructRAGModel=Mistral-7B, Retriever=Contriever, Venue=ICLR ’252026.05 | 25 | — | — | — | — | — | — | — | 15 | 51 | 28 | 32 | |
| InstructRAGModel=Mistral-7B, Retriever=DPR, Venue=ICLR ’252026.05 | 25 | — | — | — | — | — | — | — | 15 | 49 | 37 | 28 | |
| InstructRAGModel=Mistral-7B, Retriever=ANCE, Venue=ICLR ’252026.05 | 24 | — | — | — | — | — | — | — | 14 | 57 | 30 | 21 | |
| RobustRAGModel=Mistral-7B, Retriever=Contriever, Venue=arXiv ’242026.05 | 24 | — | — | — | — | — | — | — | 4 | 2 | 28 | 11 | |
| RobustRAGModel=Mistral-7B, Retriever=DPR, Venue=arXiv ’242026.05 | 23 | — | — | — | — | — | — | — | 3 | 8 | 18 | 13 | |
| RobustRAGModel=Mistral-7B, Retriever=ANCE, Venue=arXiv ’242026.05 | 20 | — | — | — | — | — | — | — | 4 | 4 | 22 | 8 | |
| ReliabilityRAGModel=Mistral-7B, Retriever=ANCE, Venue=NeurIPS ’252026.05 | 18 | — | — | — | — | — | — | — | 3 | 6 | 20 | 5 | |
| ReliabilityRAGModel=Mistral-7B, Retriever=Contriever, Venue=NeurIPS ’252026.05 | 18 | — | — | — | — | — | — | — | 3 | 6 | 19 | 16 | |
| ReliabilityRAGModel=Mistral-7B, Retriever=DPR, Venue=NeurIPS ’252026.05 | 12 | — | — | — | — | — | — | — | 3 | 11 | 17 | 12 | |
| Corpus Poisoning Attack2026.04 | — | 51 | 0 | 0 | 8,247.5 | 100 | 9.16 | 29 | — | — | — | — | |
| PoisonedRAG (B)2026.04 | — | 87 | 60 | 67 | 55.88 | 28 | 2.21 | 100 | — | — | — | — | |
| PoisonedRAG (W)2026.04 | — | 95 | 63 | 66 | 257.54 | 0 | 6.02 | 65 | — | — | — | — | |
| Prompt Injection Attack2026.04 | — | 77 | 83 | 81 | 137.86 | 100 | 0.99 | 100 | — | — | — | — | |
| RefineRAG2026.04 | — | 70 | 83 | 81 | 127.53 | 1 | 0.86 | 53 | — | — | — | — |