Multimodal Retrieval-Augmented Generation on MMQA
53.6ROrig.LPA-BB
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| LPA-BBGenerator=LLaVA, Retriever=CLIP-ViT-L, Retriever config (Rt.)=N = m, Reranker config (Rr.)=✗, Captions=-2025.02 | 53.6 | 41.6 | 36 | 22.4 | |
| LPA-BBGenerator=Qwen-VL-Chat, Retriever=CLIP-ViT-L, Retriever config (Rt.)=N = m, Reranker config (Rr.)=✗, Captions=-2025.02 | 53.6 | 40 | 36 | 26.4 | |
| LPA-BBGenerator=LLaVA, Retriever=CLIP-ViT-L, Retriever config (Rt.)=N = 5, Reranker config (Rr.)=K = m, Captions=✗2025.02 | 40.8 | 33.6 | 43.2 | 36.8 | |
| LPA-BBGenerator=LLaVA, Retriever=CLIP-ViT-L, Retriever config (Rt.)=N = 5, Reranker config (Rr.)=K = m, Captions=✓2025.02 | 37.6 | 33.6 | 55.2 | 40 | |
| LPA-BBGenerator=Qwen-VL-Chat, Retriever=CLIP-ViT-L, Retriever config (Rt.)=N = 5, Reranker config (Rr.)=K = m, Captions=✗2025.02 | 36.8 | 31.2 | 49.6 | 38.4 | |
| LPA-RtGenerator=Qwen-VL-Chat, Retriever=CLIP-ViT-L, Retriever config (Rt.)=N = 5, Reranker config (Rr.)=K = m, Captions=✗2025.02 | 35.2 | 27.2 | 52 | 38.4 | |
| LPA-RtGenerator=LLaVA, Retriever=CLIP-ViT-L, Retriever config (Rt.)=N = 5, Reranker config (Rr.)=K = m, Captions=✗2025.02 | 28 | 23.2 | 60.8 | 47.2 | |
| LPA-BBGenerator=Qwen-VL-Chat, Retriever=CLIP-ViT-L, Retriever config (Rt.)=N = 5, Reranker config (Rr.)=K = m, Captions=✓2025.02 | 26.4 | 24.8 | 68.8 | 46.4 | |
| LPA-RtGenerator=LLaVA, Retriever=CLIP-ViT-L, Retriever config (Rt.)=N = 5, Reranker config (Rr.)=K = m, Captions=✓2025.02 | 23.2 | 19.2 | 74.4 | 48.8 | |
| LPA-RtGenerator=Qwen-VL-Chat, Retriever=CLIP-ViT-L, Retriever config (Rt.)=N = 5, Reranker config (Rr.)=K = m, Captions=✓2025.02 | 22.4 | 20.8 | 75.2 | 49.6 | |
| LPA-RtGenerator=LLaVA, Retriever=CLIP-ViT-L, Retriever config (Rt.)=N = m, Reranker config (Rr.)=✗, Captions=-2025.02 | 8.8 | 11.2 | 88.8 | 56.8 | |
| LPA-RtGenerator=Qwen-VL-Chat, Retriever=CLIP-ViT-L, Retriever config (Rt.)=N = m, Reranker config (Rr.)=✗, Captions=-2025.02 | 8.8 | 12 | 88.8 | 55.2 |