Question Answering on HotpotQA PIA (test)
90.2ASRVanilla
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| VanillaLLM=Llama 2-C2025.06 | 90.2 | — | 3.6 | |
| VanillaLLM=GPT-4o2025.06 | 78.8 | — | 10.6 | |
| RobustRAGLLM Backbone=Llama-3.1-8B, Retrieval Model=Contriever2026.05 | 78 | 52 | — | |
| RobustRAGLLM Backbone=Qwen2.5-7B, Retrieval Model=ANCE2026.05 | 74 | 50 | — | |
| RobustRAGLLM Backbone=Qwen2.5-7B, Retrieval Model=Contriever2026.05 | 73 | 52 | — | |
| RobustRAGLLM Backbone=Llama-3.1-8B, Retrieval Model=ANCE2026.05 | 73 | 42 | — | |
| RobustRAGLLM Backbone=Llama-3.1-8B, Retrieval Model=DPR2026.05 | 70 | 42 | — | |
| VanillaLLM=Mistral-7B2025.06 | 69 | — | 18.6 | |
| InstructRAGLLM Backbone=Mistral-7B, Retrieval Model=ANCE2026.05 | 67 | 29 | — | |
| RobustRAGLLM Backbone=Qwen2.5-7B, Retrieval Model=DPR2026.05 | 65 | 41 | — | |
| InstructRAGLLM Backbone=Mistral-7B, Retrieval Model=DPR2026.05 | 65 | 29 | — | |
| InstructRAGLLM Backbone=Qwen2.5-7B, Retrieval Model=DPR2026.05 | 63 | 53 | — | |
| InstructRAGLLM Backbone=Qwen2.5-7B, Retrieval Model=ANCE2026.05 | 60 | 49 | — | |
| ReliabilityRAGLLM Backbone=Llama-3.1-8B, Retrieval Model=DPR2026.05 | 60 | 33 | — | |
| InstructRAGLLM Backbone=Qwen2.5-7B, Retrieval Model=Contriever2026.05 | 57 | 59 | — | |
| InstructRAGLLM Backbone=Mistral-7B, Retrieval Model=Contriever2026.05 | 57 | 40 | — | |
| InstructRAGLLM Backbone=Llama-3.1-8B, Retrieval Model=ANCE2026.05 | 53 | 50 | — | |
| ReliabilityRAGLLM Backbone=Llama-3.1-8B, Retrieval Model=Contriever2026.05 | 51 | 40 | — | |
| BiRDLLM Backbone=Llama-3.1-8B, Retrieval Model=DPR2026.05 | 50 | 64 | — | |
| ReliabilityRAGLLM Backbone=Llama-3.1-8B, Retrieval Model=ANCE2026.05 | 49 | 31 | — | |
| BiRDLLM Backbone=Qwen2.5-7B, Retrieval Model=DPR2026.05 | 48 | 54 | — | |
| InstructRAGLLM Backbone=Llama-3.1-8B, Retrieval Model=Contriever2026.05 | 44 | 42 | — | |
| AV FilterLLM=Llama 2-C, alpha=∞2025.06 | 44 | — | 17.8 | |
| VanillaRAGLLM Backbone=Llama-3.1-8B, Retrieval Model=DPR2026.05 | 43 | 56 | — | |
| ReliabilityRAGLLM Backbone=Qwen2.5-7B, Retrieval Model=ANCE2026.05 | 42 | 33 | — | |
| VanillaRAGLLM Backbone=Qwen2.5-7B, Retrieval Model=DPR2026.05 | 42 | 55 | — | |
| ReliabilityRAGLLM Backbone=Qwen2.5-7B, Retrieval Model=DPR2026.05 | 42 | 39 | — | |
| VanillaRAGLLM Backbone=Mistral-7B, Retrieval Model=DPR2026.05 | 39 | 46 | — | |
| BiRDLLM Backbone=Llama-3.1-8B, Retrieval Model=ANCE2026.05 | 39 | 64 | — | |
| BiRDLLM Backbone=Mistral-7B, Retrieval Model=DPR2026.05 | 38 | 48 | — | |
| InstructRAGLLM Backbone=Llama-3.1-8B, Retrieval Model=DPR2026.05 | 38 | 35 | — | |
| BiRDLLM Backbone=Llama-3.1-8B, Retrieval Model=Contriever2026.05 | 37 | 72 | — | |
| ReliabilityRAGLLM Backbone=Qwen2.5-7B, Retrieval Model=Contriever2026.05 | 36 | 39 | — | |
| BiRDLLM Backbone=Qwen2.5-7B, Retrieval Model=Contriever2026.05 | 36 | 60 | — | |
| VanillaRAGLLM Backbone=Llama-3.1-8B, Retrieval Model=ANCE2026.05 | 35 | 60 | — | |
| VanillaRAGLLM Backbone=Qwen2.5-7B, Retrieval Model=ANCE2026.05 | 34 | 49 | — | |
| BiRDLLM Backbone=Qwen2.5-7B, Retrieval Model=ANCE2026.05 | 34 | 49 | — | |
| VanillaRAGLLM Backbone=Llama-3.1-8B, Retrieval Model=Contriever2026.05 | 33 | 69 | — | |
| DecodingLLM=Llama 2-C2025.06 | 33 | — | 23 | |
| VanillaRAGLLM Backbone=Qwen2.5-7B, Retrieval Model=Contriever2026.05 | 32 | 54 | — | |
| DecodingLLM=Mistral-7B2025.06 | 28 | — | 38 | |
| VanillaRAGLLM Backbone=Mistral-7B, Retrieval Model=ANCE2026.05 | 27 | 49 | — | |
| BiRDLLM Backbone=Mistral-7B, Retrieval Model=ANCE2026.05 | 27 | 46 | — | |
| VanillaRAGLLM Backbone=Mistral-7B, Retrieval Model=Contriever2026.05 | 27 | 67 | — | |
| BiRDLLM Backbone=Mistral-7B, Retrieval Model=Contriever2026.05 | 27 | 59 | — | |
| KeywordLLM=Llama 2-C2025.06 | 25 | — | 36 | |
| RobustRAGLLM Backbone=Mistral-7B, Retrieval Model=Contriever2026.05 | 22 | 6 | — | |
| KeywordLLM=Mistral-7B2025.06 | 21 | — | 48 | |
| RobustRAGLLM Backbone=Mistral-7B, Retrieval Model=ANCE2026.05 | 19 | 4 | — | |
| ReliabilityRAGLLM Backbone=Mistral-7B, Retrieval Model=ANCE2026.05 | 19 | 4 | — | |
| ReliabilityRAGLLM Backbone=Mistral-7B, Retrieval Model=Contriever2026.05 | 18 | 4 | — | |
| AV FilterLLM=GPT-4o, alpha=∞2025.06 | 17.8 | — | 35.6 | |
| KeywordLLM=GPT-4o2025.06 | 17.4 | — | 43.6 | |
| RobustRAGLLM Backbone=Mistral-7B, Retrieval Model=DPR2026.05 | 17 | 5 | — | |
| AV FilterLLM=Mistral-7B, alpha=∞2025.06 | 13.4 | — | 47.2 | |
| ReliabilityRAGLLM Backbone=Mistral-7B, Retrieval Model=DPR2026.05 | 13 | 4 | — | |
| AV FilterLLM=Llama 2-C, alpha=52025.06 | 11.4 | — | 34 | |
| AV FilterLLM=GPT-4o, alpha=102025.06 | 11.2 | — | 40 | |
| AV FilterLLM=Llama 2-C, alpha=102025.06 | 10.4 | — | 34.4 | |
| AV FilterLLM=GPT-4o, alpha=52025.06 | 9.8 | — | 42.6 | |
| AV FilterLLM=Mistral-7B, alpha=102025.06 | 8.4 | — | 52.6 | |
| AV FilterLLM=Mistral-7B, alpha=52025.06 | 8 | — | 53 |