Visual Question Answering on InfoSeek (Full)
43AccuracyCgRAG
Evaluation results
| Method | Links | |
|---|---|---|
| CgRAGLLM/MLLM=InternVL3-8B, Retriever=EVA-CLIP-8B, Feature=V+T2026.05 | 43 | |
| VLM-PRF w/ RLLLM/MLLM=InternVL3-8B, Retriever=EVA-CLIP-8B, Feature=V+T2026.05 | 42.5 | |
| CgRAGLLM/MLLM=LLaVA-NeXT-8B, Retriever=EVA-CLIP-8B, Feature=V+T2026.05 | 39.7 | |
| VLM-PRFLLM/MLLM=Qwen2.5-VL-7B, Retriever=EVA-CLIP-8B, Feature=V+T2026.05 | 39.5 | |
| CgRAGLLM/MLLM=LLaVA-NeXT-7B, Retriever=EVA-CLIP-8B, Feature=V+T2026.05 | 39.5 | |
| CgRAGLLM/MLLM=Qwen2.5-VL-7B, Retriever=EVA-CLIP-8B, Feature=V+T2026.05 | 36.6 | |
| MMKB-RAGLLM/MLLM=Qwen2-VL-7B, Retriever=EVA-CLIP-8B, Feature=V2026.05 | 36.4 | |
| CgRAGLLM/MLLM=Qwen2-VL-7B, Retriever=EVA-CLIP-8B, Feature=V+T2026.05 | 36.3 | |
| Wiki-LLaVALLM/MLLM=Vicuna-7B, Retriever=CLIP ViT-L/14 + Contriever, Feature=T2026.05 | 28.9 | |
| ReflectiVALLM/MLLM=LLaMA-3.1-8B, Retriever=EVA-CLIP-8B, Feature=V2026.05 | 28.3 | |
| EchoSightLLM/MLLM=Mistral-7B/LLaMA-3-8B, Retriever=EVA-CLIP-8B, Feature=V2026.05 | 27.5 | |
| Wiki-LLaVALLM/MLLM=LLaMA-3.1-8B, Retriever=CLIP ViT-L/14 + Contriever, Feature=T2026.05 | 27.1 | |
| KU-RAGLLM/MLLM=GPT-4o, Retriever=EVA-CLIP-8B, Feature=V+T2026.05 | 26.1 | |
| RORA-VLMLLM/MLLM=Vicuna-7B, Retriever=CLIP + Google Search, Feature=V2026.05 | 26 | |
| LLaVA-NeXT-8BMode=Zero-shot MLLMs2026.05 | 23.6 | |
| Qwen2.5-VL-7BMode=Zero-shot MLLMs2026.05 | 23.2 | |
| LLaVA-NeXT-7BMode=Zero-shot MLLMs2026.05 | 23.1 | |
| RMCDBase LVLM=InternVL 2.5 (38B)2026.01 | 22.9 | |
| ConcatBase LVLM=InternVL 2.5 (38B)2026.01 | 22.3 | |
| RMCDBase LVLM=BLIP-2 (T5-XXL)2026.01 | 21 | |
| RAGBase LVLM=InternVL 2.5 (38B)2026.01 | 20.7 | |
| Qwen2.5-VL-3BMode=Zero-shot MLLMs2026.05 | 20.7 | |
| SCDBase LVLM=InternVL 2.5 (38B)2026.01 | 20.4 | |
| SCDBase LVLM=BLIP-2 (T5-XXL)2026.01 | 19.6 | |
| SCDBase LVLM=LLaVA-1.5 (Vicuna 13B)2026.01 | 19.6 | |
| RAGBase LVLM=BLIP-2 (T5-XXL)2026.01 | 19.3 | |
| RMCDBase LVLM=LLaVA-1.5 (Vicuna 13B)2026.01 | 19.2 | |
| RAGBase LVLM=LLaVA-1.5 (Vicuna 13B)2026.01 | 18.9 | |
| EchoSightLLM/MLLM=LLaMA-3.1-8B, Retriever=EVA-CLIP-8B, Feature=V2026.05 | 18.8 | |
| ConcatBase LVLM=BLIP-2 (T5-XXL)2026.01 | 18.7 | |
| RMCDBase LVLM=BLIP-2 (T5-XL)2026.01 | 18.1 | |
| RAGBase LVLM=BLIP-2 (T5-XL)2026.01 | 17.4 | |
| Max ProbabilityBase LVLM=InternVL 2.5 (38B)2026.01 | 17.2 | |
| SCDBase LVLM=BLIP-2 (T5-XL)2026.01 | 17.1 | |
| Max ProbabilityBase LVLM=LLaVA-1.5 (Vicuna 13B)2026.01 | 16.9 | |
| Qwen2-VL-7BMode=Zero-shot MLLMs2026.05 | 16.1 | |
| Max ProbabilityBase LVLM=BLIP-2 (T5-XXL)2026.01 | 15.5 | |
| ConcatBase LVLM=LLaVA-1.5 (Vicuna 13B)2026.01 | 14.9 | |
| Max ProbabilityBase LVLM=BLIP-2 (T5-XL)2026.01 | 14.8 | |
| GPT-4VMode=Zero-shot MLLMs2026.05 | 14.6 | |
| ConsistencyBase LVLM=InternVL 2.5 (38B)2026.01 | 13.8 | |
| UnconditionalBase LVLM=InternVL 2.5 (38B)2026.01 | 13.3 | |
| BLIP-2LLM/MLLM=Flan-T5XL, Mode=Zero-shot MLLMs2026.05 | 12.5 | |
| DPR_V+TLLM/MLLM=Multi-passage BERT, Retriever=CLIP ViT-B/32, Feature=V+T2026.05 | 12.4 | |
| ConcatBase LVLM=BLIP-2 (T5-XL)2026.01 | 12.1 | |
| ConsistencyBase LVLM=BLIP-2 (T5-XXL)2026.01 | 12 | |
| ConsistencyBase LVLM=LLaVA-1.5 (Vicuna 13B)2026.01 | 12 | |
| ConsistencyBase LVLM=BLIP-2 (T5-XL)2026.01 | 9.6 | |
| LLaVA-1.5LLM/MLLM=Vicuna-7B, Mode=Zero-shot MLLMs2026.05 | 9.5 | |
| InstructBLIPLLM/MLLM=Flan-T5XL, Mode=Zero-shot MLLMs2026.05 | 8.1 | |
| UnconditionalBase LVLM=BLIP-2 (T5-XXL)2026.01 | 7.9 | |
| LLaVA-1.5LLM/MLLM=LLaMA-3.1-8B, Mode=Zero-shot MLLMs2026.05 | 7.8 | |
| UnconditionalBase LVLM=LLaVA-1.5 (Vicuna 13B)2026.01 | 7.7 | |
| UnconditionalBase LVLM=BLIP-2 (T5-XL)2026.01 | 7.3 | |
| RMCDBase LVLM=BLIP-2 (OPT 6.7B)2026.01 | 5.8 | |
| SCDBase LVLM=BLIP-2 (OPT 6.7B)2026.01 | 4.9 | |
| UnconditionalBase LVLM=BLIP-2 (OPT 6.7B)2026.01 | 4 | |
| RAGBase LVLM=BLIP-2 (OPT 6.7B)2026.01 | 4 | |
| ConsistencyBase LVLM=BLIP-2 (OPT 6.7B)2026.01 | 1.2 | |
| Max ProbabilityBase LVLM=BLIP-2 (OPT 6.7B)2026.01 | 1.2 | |
| ConcatBase LVLM=BLIP-2 (OPT 6.7B)2026.01 | 0.5 |