Question Answering on HotpotQA PoisonedRAG (test)
17Abstention Rate (ASR)BiRD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| BiRDLLM Backbone=Mistral-7B, Retrieval Model=Contriever2026.05 | 17 | 68 | |
| BiRDLLM Backbone=Mistral-7B, Retrieval Model=ANCE2026.05 | 18 | 67 | |
| BiRDLLM Backbone=Qwen2.5-7B, Retrieval Model=ANCE2026.05 | 22 | 59 | |
| BiRDLLM Backbone=Qwen2.5-7B, Retrieval Model=Contriever2026.05 | 24 | 69 | |
| BiRDLLM Backbone=Mistral-7B, Retrieval Model=DPR2026.05 | 28 | 59 | |
| RobustRAGLLM Backbone=Mistral-7B, Retrieval Model=Contriever2026.05 | 29 | 10 | |
| ReliabilityRAGLLM Backbone=Mistral-7B, Retrieval Model=ANCE2026.05 | 33 | 12 | |
| BiRDLLM Backbone=Llama-3.1-8B, Retrieval Model=ANCE2026.05 | 33 | 82 | |
| BiRDLLM Backbone=Llama-3.1-8B, Retrieval Model=Contriever2026.05 | 33 | 85 | |
| RobustRAGLLM Backbone=Mistral-7B, Retrieval Model=DPR2026.05 | 35 | 12 | |
| ReliabilityRAGLLM Backbone=Mistral-7B, Retrieval Model=DPR2026.05 | 35 | 11 | |
| ReliabilityRAGLLM Backbone=Mistral-7B, Retrieval Model=Contriever2026.05 | 36 | 13 | |
| ReliabilityRAGLLM Backbone=Llama-3.1-8B, Retrieval Model=ANCE2026.05 | 36 | 30 | |
| ReliabilityRAGLLM Backbone=Llama-3.1-8B, Retrieval Model=Contriever2026.05 | 38 | 29 | |
| BiRDLLM Backbone=Qwen2.5-7B, Retrieval Model=DPR2026.05 | 41 | 62 | |
| RobustRAGLLM Backbone=Mistral-7B, Retrieval Model=ANCE2026.05 | 41 | 11 | |
| RobustRAGLLM Backbone=Llama-3.1-8B, Retrieval Model=ANCE2026.05 | 44 | 37 | |
| ReliabilityRAGLLM Backbone=Llama-3.1-8B, Retrieval Model=DPR2026.05 | 44 | 26 | |
| RobustRAGLLM Backbone=Llama-3.1-8B, Retrieval Model=Contriever2026.05 | 45 | 37 | |
| RobustRAGLLM Backbone=Llama-3.1-8B, Retrieval Model=DPR2026.05 | 51 | 38 | |
| BiRDLLM Backbone=Llama-3.1-8B, Retrieval Model=DPR2026.05 | 51 | 78 | |
| ReliabilityRAGLLM Backbone=Qwen2.5-7B, Retrieval Model=DPR2026.05 | 60 | 31 | |
| InstructRAGLLM Backbone=Llama-3.1-8B, Retrieval Model=Contriever2026.05 | 61 | 51 | |
| ReliabilityRAGLLM Backbone=Qwen2.5-7B, Retrieval Model=ANCE2026.05 | 65 | 36 | |
| InstructRAGLLM Backbone=Llama-3.1-8B, Retrieval Model=DPR2026.05 | 68 | 41 | |
| ReliabilityRAGLLM Backbone=Qwen2.5-7B, Retrieval Model=Contriever2026.05 | 70 | 39 | |
| RobustRAGLLM Backbone=Qwen2.5-7B, Retrieval Model=DPR2026.05 | 78 | 30 | |
| InstructRAGLLM Backbone=Mistral-7B, Retrieval Model=ANCE2026.05 | 78 | 24 | |
| RobustRAGLLM Backbone=Qwen2.5-7B, Retrieval Model=Contriever2026.05 | 79 | 41 | |
| InstructRAGLLM Backbone=Llama-3.1-8B, Retrieval Model=ANCE2026.05 | 79 | 42 | |
| VanillaRAGLLM Backbone=Llama-3.1-8B, Retrieval Model=Contriever2026.05 | 79 | 71 | |
| InstructRAGLLM Backbone=Mistral-7B, Retrieval Model=Contriever2026.05 | 80 | 28 | |
| InstructRAGLLM Backbone=Qwen2.5-7B, Retrieval Model=ANCE2026.05 | 81 | 20 | |
| InstructRAGLLM Backbone=Qwen2.5-7B, Retrieval Model=Contriever2026.05 | 82 | 16 | |
| InstructRAGLLM Backbone=Qwen2.5-7B, Retrieval Model=DPR2026.05 | 84 | 19 | |
| RobustRAGLLM Backbone=Qwen2.5-7B, Retrieval Model=ANCE2026.05 | 88 | 33 | |
| VanillaRAGLLM Backbone=Llama-3.1-8B, Retrieval Model=ANCE2026.05 | 88 | 65 | |
| InstructRAGLLM Backbone=Mistral-7B, Retrieval Model=DPR2026.05 | 89 | 19 | |
| VanillaRAGLLM Backbone=Llama-3.1-8B, Retrieval Model=DPR2026.05 | 90 | 61 | |
| VanillaRAGLLM Backbone=Qwen2.5-7B, Retrieval Model=ANCE2026.05 | 93 | 24 | |
| VanillaRAGLLM Backbone=Qwen2.5-7B, Retrieval Model=Contriever2026.05 | 93 | 29 | |
| VanillaRAGLLM Backbone=Mistral-7B, Retrieval Model=ANCE2026.05 | 93 | 15 | |
| VanillaRAGLLM Backbone=Mistral-7B, Retrieval Model=Contriever2026.05 | 93 | 16 | |
| VanillaRAGLLM Backbone=Qwen2.5-7B, Retrieval Model=DPR2026.05 | 95 | 24 | |
| VanillaRAGLLM Backbone=Mistral-7B, Retrieval Model=DPR2026.05 | 95 | 13 |