Multimodal Question Answering on MMQA k=1
57.4ROrig@kCLIP + BLIP-2
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| CLIP + BLIP-2Retriever=CLIP, Generator=BLIP-22026.02 | 57.4 | 40.4 | 31.3 | 50 | |
| CLIP + LLaVARetriever=CLIP, Generator=LLaVA2026.02 | 57.4 | 40.4 | 41.3 | 40.1 | |
| OpenCLIP + BLIP-2Retriever=OpenCLIP, Generator=BLIP-22026.02 | 55.7 | 43.5 | 29.1 | 53 | |
| OpenCLIP + LLaVARetriever=OpenCLIP, Generator=LLaVA2026.02 | 55.7 | 43.5 | 38.7 | 40.1 | |
| FLAVA + BLIP-2Retriever=FLAVA, Generator=BLIP-22026.02 | 44.3 | 38.3 | 26.1 | 48 | |
| FLAVA + LLaVARetriever=FLAVA, Generator=LLaVA2026.02 | 44.3 | 38.3 | 34.3 | 36.6 | |
| GPA-RtRrGenRetriever=CLIP-ViT-L, Reranker/Generator=LLaVA, N=5, K=m, Captioning=false2025.02 | 30.4 | — | 23.2 | — | |
| GPA-RtRrGenRetriever=CLIP-ViT-L, Reranker/Generator=Qwen-VL-Chat, N=5, K=m, Captioning=true2025.02 | 23.2 | — | 12.8 | — | |
| GPA-RtRrGenRetriever=CLIP-ViT-L, Reranker/Generator=LLaVA, N=5, K=m, Captioning=true2025.02 | 17.6 | — | 18.4 | — | |
| GPA-RtRrGenRetriever=CLIP-ViT-L, Reranker/Generator=Qwen-VL-Chat, N=5, K=m, Captioning=false2025.02 | 6.4 | — | 3.2 | — | |
| GPA-RtRrGenRetriever=CLIP-ViT-L, Reranker/Generator=LLaVA, N=m, Captioning=false2025.02 | 5.6 | — | 9.6 | — | |
| GPA-RtRrGenRetriever=CLIP-ViT-L, Reranker/Generator=Qwen-VL-Chat, N=m, Captioning=false2025.02 | 2.4 | — | 1.6 | — | |
| GPA-RtRetriever=CLIP-ViT-L, Reranker/Generator=LLaVA, N=m, Captioning=false2025.02 | 1.6 | — | 8.8 | — | |
| GPA-RtRetriever=CLIP-ViT-L, Reranker/Generator=LLaVA, N=5, K=m, Captioning=false2025.02 | 1.6 | — | 8.8 | — | |
| GPA-RtRetriever=CLIP-ViT-L, Reranker/Generator=LLaVA, N=5, K=m, Captioning=true2025.02 | 1.6 | — | 8.8 | — | |
| GPA-RtRetriever=CLIP-ViT-L, Reranker/Generator=Qwen-VL-Chat, N=m, Captioning=false2025.02 | 1.6 | — | 8.8 | — | |
| GPA-RtRetriever=CLIP-ViT-L, Reranker/Generator=Qwen-VL-Chat, N=5, K=m, Captioning=false2025.02 | 1.6 | — | 8.8 | — | |
| GPA-RtRetriever=CLIP-ViT-L, Reranker/Generator=Qwen-VL-Chat, N=5, K=m, Captioning=true2025.02 | 1.6 | — | 8.8 | — | |
| SigLIP + BLIP-2Retriever=SigLIP, Generator=BLIP-22026.02 | 1.3 | 86.5 | 7.4 | 91.6 | |
| SigLIP + LLaVARetriever=SigLIP, Generator=LLaVA2026.02 | 1.3 | 86.5 | 11.3 | 73.3 |