Visual Question Answering on E-VQA (test)
84.03AccuracyMAD-RAG
Evaluation Results
| Method | Links | |
|---|---|---|
| MAD-RAGModel Family=Qwen3-VL-8B2026.01 | 84.03 | |
| MAD-RAGModel Family=Qwen3-VL-4B2026.01 | 83.84 | |
| MAD-RAGBackbone=Qwen2.5-VL-7B, Context Setting=Oracle Contexts2026.01 | 83.31 | |
| ALFARBackbone=Qwen2.5-VL-7B, Context Setting=Oracle Contexts2026.01 | 82.4 | |
| Vanilla RAGBackbone=Qwen2.5-VL-7B, Context Setting=Oracle Contexts2026.01 | 82.21 | |
| MAD-RAGBackbone=Qwen2.5-VL-3B, Context Setting=Oracle Contexts2026.01 | 80.11 | |
| CADBackbone=Qwen2.5-VL-7B, Context Setting=Oracle Contexts2026.01 | 77.89 | |
| Vanilla RAGModel Family=Qwen3-VL-8B2026.01 | 77.6 | |
| ALFARBackbone=Qwen2.5-VL-3B, Context Setting=Oracle Contexts2026.01 | 75.97 | |
| AdaCADBackbone=Qwen2.5-VL-7B, Context Setting=Oracle Contexts2026.01 | 75.65 | |
| Vanilla RAGBackbone=Qwen2.5-VL-3B, Context Setting=Oracle Contexts2026.01 | 75.63 | |
| DoLABackbone=Qwen2.5-VL-7B, Context Setting=Oracle Contexts2026.01 | 75.5 | |
| Vanilla RAGModel Family=Qwen3-VL-4B2026.01 | 75.49 | |
| VCDBackbone=Qwen2.5-VL-7B, Context Setting=Oracle Contexts2026.01 | 74.99 | |
| CADBackbone=Qwen2.5-VL-3B, Context Setting=Oracle Contexts2026.01 | 74.11 | |
| AdaCADBackbone=Qwen2.5-VL-3B, Context Setting=Oracle Contexts2026.01 | 73.04 | |
| DoLABackbone=Qwen2.5-VL-3B, Context Setting=Oracle Contexts2026.01 | 72.4 | |
| VCDBackbone=Qwen2.5-VL-3B, Context Setting=Oracle Contexts2026.01 | 71.57 | |
| MAD-RAGBackbone=LLaVA-1.5-13B, Context Setting=Oracle Contexts2026.01 | 69.07 | |
| VHRBackbone=Qwen2.5-VL-3B, Context Setting=Oracle Contexts2026.01 | 63.52 | |
| MAD-RAGBackbone=LLaVA-1.5-7B, Context Setting=Oracle Contexts2026.01 | 63.09 | |
| Vanilla RAGBackbone=LLaVA-1.5-13B, Context Setting=Oracle Contexts2026.01 | 61.23 | |
| ALFARBackbone=LLaVA-1.5-13B, Context Setting=Oracle Contexts2026.01 | 60.36 | |
| CADBackbone=LLaVA-1.5-13B, Context Setting=Oracle Contexts2026.01 | 60.19 | |
| AdaCADBackbone=LLaVA-1.5-13B, Context Setting=Oracle Contexts2026.01 | 59.47 | |
| VHRBackbone=LLaVA-1.5-13B, Context Setting=Oracle Contexts2026.01 | 58.88 | |
| VHRBackbone=Qwen2.5-VL-7B, Context Setting=Oracle Contexts2026.01 | 58.67 | |
| DoLABackbone=LLaVA-1.5-13B, Context Setting=Oracle Contexts2026.01 | 58.21 | |
| V-Retrver-7BModel Scale=7B2026.02 | 58 | |
| SPINBackbone=LLaVA-1.5-13B, Context Setting=Oracle Contexts2026.01 | 57.95 | |
| VCDBackbone=LLaVA-1.5-13B, Context Setting=Oracle Contexts2026.01 | 57.92 | |
| OPERABackbone=LLaVA-1.5-13B, Context Setting=Oracle Contexts2026.01 | 57.47 | |
| LamRA-7BModel Scale=7B2026.02 | 56.2 | |
| WikiSeekerGenerator Model=Inspector+Qwen2.5-7B-Instruct, Gen. FT=true, Ret. FT=true2026.04 | 55.62 | |
| CADBackbone=LLaVA-1.5-7B, Context Setting=Oracle Contexts2026.01 | 54.88 | |
| AdaCADBackbone=LLaVA-1.5-7B, Context Setting=Oracle Contexts2026.01 | 54.69 | |
| RA-VQAv2 w/ PreFLMRRAG System=RA-VQAv2, Retriever=PreFLMR2026.02 | 54.5 | |
| Vanilla RAGBackbone=LLaVA-1.5-7B, Context Setting=Oracle Contexts2026.01 | 53.89 | |
| ALFARBackbone=LLaVA-1.5-7B, Context Setting=Oracle Contexts2026.01 | 53.2 | |
| VCDBackbone=LLaVA-1.5-7B, Context Setting=Oracle Contexts2026.01 | 53.2 | |
| VHRBackbone=LLaVA-1.5-7B, Context Setting=Oracle Contexts2026.01 | 53.04 | |
| WikiSeekerGenerator Model=Qwen2.5-7B-Instruct, Gen. FT=false, Ret. FT=true2026.04 | 52.97 | |
| WikiSeekerGenerator Model=Qwen3-VL-8B-Instruct, Gen. FT=false, Ret. FT=true2026.04 | 51.81 | |
| DoLABackbone=LLaVA-1.5-7B, Context Setting=Oracle Contexts2026.01 | 51.01 | |
| OMGMGenerator Model=LLaVA-1.5-7B, Gen. FT=false, Ret. FT=true2025.05 | 50.17 | |
| OMGMGenerator Model=LLaVA-1.5-7B, Gen. FT=true, Ret. FT=true2026.04 | 50.17 | |
| OMGMGenerator Model=LLaMA3-8B, Gen. FT=false, Ret. FT=true2025.05 | 49.94 | |
| OMGMGenerator Model=InternVL-2.5-8B, Gen. FT=false, Ret. FT=true2025.05 | 48.72 | |
| SPINBackbone=LLaVA-1.5-7B, Context Setting=Oracle Contexts2026.01 | 48.72 | |
| OPERABackbone=LLaVA-1.5-7B, Context Setting=Oracle Contexts2026.01 | 48.56 | |
| EchoSightGenerator Model=Mistral-7B | LLaMA3-8B, Gen. FT=false, Ret. FT=true2025.05 | 41.8 | |
| EchoSightGenerator Model=Mistral-7B | LLaMA3-8B, Gen. FT=false, Ret. FT=true2026.04 | 41.8 | |
| ReflectiVAGenerator Model=LLAVA-MORE-8B, Gen. FT=true, Ret. FT=true2025.05 | 35.5 | |
| ReflectiVAGenerator Model=LLaVA-MORE-8B, Gen. FT=true, Ret. FT=true2026.04 | 35.5 | |
| Closed-bookModel Family=Qwen3-VL-8B2026.01 | 27.01 | |
| Closed-book (parametric)Backbone=Qwen2.5-VL-7B, Context Setting=Parametric only2026.01 | 25.15 | |
| Closed-bookModel Family=Qwen3-VL-4B2026.01 | 23.04 | |
| Closed-book (parametric)Backbone=Qwen2.5-VL-3B, Context Setting=Parametric only2026.01 | 22.93 | |
| Wiki-LLaVAGenerator Model=LLaVA-1.5-7B, Gen. FT=false, Ret. FT=true2025.05 | 21.8 | |
| Wiki-LLaVAGenerator Model=LLaVA-1.5-7B, Gen. FT=true, Ret. FT=false2026.04 | 21.8 | |
| RORA-VLMGenerator Model=LLaVA-1.5-7B, Gen. FT=false, Ret. FT=true2025.05 | 20.29 | |
| RoRA-VLMGenerator Model=LLaVA-1.5-7B, Gen. FT=true, Ret. FT=false2026.04 | 20.29 | |
| MMA-RAGModel=Qwen2VL-7B2026.02 | 20 | |
| RIRModel=Qwen2VL-7B2026.02 | 19.8 | |
| P(true)Model=Qwen2VL-7B2026.02 | 19.1 | |
| Closed-book (parametric)Backbone=LLaVA-1.5-13B, Context Setting=Parametric only2026.01 | 18.8 | |
| CLIPModel=Qwen2VL-7B2026.02 | 18.5 | |
| Closed-book (parametric)Backbone=LLaVA-1.5-7B, Context Setting=Parametric only2026.01 | 17.57 | |
| MMA-RAGModel=Idefics2-8B2026.02 | 14.6 | |
| CoTModel=Qwen2VL-7B2026.02 | 14.4 | |
| CLIPModel=Idefics2-8B2026.02 | 14.3 | |
| Few shotModel=Qwen2VL-7B2026.02 | 14.1 | |
| RIRModel=Idefics2-8B2026.02 | 14.1 | |
| CoTModel=Idefics2-8B2026.02 | 14.1 | |
| P(true)Model=Idefics2-8B2026.02 | 13.8 | |
| Zero shotModel=Idefics2-8B2026.02 | 13 | |
| Few shotModel=Idefics2-8B2026.02 | 12.7 | |
| MMA-RAGModel=Idefics3-8B2026.02 | 12.6 | |
| RIRModel=Idefics3-8B2026.02 | 12.5 | |
| Zero shotModel=Idefics3-8B2026.02 | 11.9 | |
| P(true)Model=Idefics3-8B2026.02 | 11 | |
| CoTModel=Idefics3-8B2026.02 | 10.9 | |
| Zero shotModel=Qwen2VL-7B2026.02 | 10 | |
| CLIPModel=Idefics3-8B2026.02 | 10 | |
| Few shotModel=Idefics3-8B2026.02 | 6.3 |