Multimodal Question Answering on WebQA k=2
64.8ROrig@kOpenCLIP + BLIP-2
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| OpenCLIP + BLIP-2Retriever=OpenCLIP, Generator=BLIP-22026.02 | 64.8 | 94.1 | 4.5 | 66.7 | |
| OpenCLIP + LLaVARetriever=OpenCLIP, Generator=LLaVA2026.02 | 64.8 | 94.1 | 6.3 | 54.7 | |
| FLAVA + BLIP-2Retriever=FLAVA, Generator=BLIP-22026.02 | 54.1 | 89.1 | 6.1 | 63.9 | |
| FLAVA + LLaVARetriever=FLAVA, Generator=LLaVA2026.02 | 54.1 | 89.1 | 7.9 | 51.6 | |
| CLIP + BLIP-2Retriever=CLIP, Generator=BLIP-22026.02 | 52.8 | 89.1 | 5.8 | 64.5 | |
| CLIP + LLaVARetriever=CLIP, Generator=LLaVA2026.02 | 52.8 | 89.1 | 7.9 | 52.4 | |
| SigLIP + BLIP-2Retriever=SigLIP, Generator=BLIP-22026.02 | 14.3 | 11.4 | 28.1 | 25.4 | |
| SigLIP + LLaVARetriever=SigLIP, Generator=LLaVA2026.02 | 14.3 | 11.4 | 29.5 | 24.2 |